This is the transparency report for Anthropic for the Claude 3 model. To see their responses for each indicator, click through the various domains and subdomains. For further information, visit the website for the May 2024 Foundation Model Transparency Index.
For the data used in building the model, is the data size disclosed?
Disclosure: Not disclosed
Note: Data size should be reported in appropriate units (e.g. bytes, words, tokens, images, frames) and broken down by modality. Data size should be reported to a precision of one significant figure (e.g. 4 trillion tokens, 200 thousand images). No form of decomposition into data phases is required.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
For all data used in building the model, are the data sources disclosed?
Disclosure: Not disclosed
Note: To receive this point, a meaningful decomposition of sources must be listed in an understandable way (e.g. named URLs/domains/databases/data providers). It does not suffice to say data is “sourced from the Internet" or comes from "licensed sources”.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
For all data used in building the model, is there some characterization of the people who created the data?
Disclosure: Not disclosed
Note: While information about data creators may not be easily discernible for some data scraped from the web, the general sources (URLs/domains) should be listed, and, for other data that is bought, licensed, or collected, a reasonable attempt at characterizing the underlying people who provided the data is required to receive this point. The relevant properties of people can vary depending on context: for example, relevant properties could include demographic information like fraction of Black individuals contributing to the dataset, geographic information like fraction of European individuals contributing to the dataset, language information like fraction of L1 English speakers, or occupational information like the fraction of professional artists.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are the selection protocols for including and excluding data sources disclosed?
Disclosure: Not disclosed
Note: Selection protocols refer to procedures used to choose which datasets or subsets of datasets will be used to build a model. We will award this point even if the selection protocols are non-exhaustive.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
For all data sources, are the curation protocols for those data sources disclosed?
Disclosure: Several data cleaning and filtering methods, including deduplication and classification
Note: Curation protocols refer to steps taken to further modify data sources, such as procedures to manage, annotate, and organize data. The aims of curation might include improving the quality, relevance, and representativeness of the data. We will award this point if the developer reports that it does not perform any further curation beyond the data sources.
References: See model card "training data" section " on page 3 of our Claude 3 model card
Justification: We award this point because we currently do not specify an explicit minimum standard for how much specificity is required in rgeards to curation practices.
New disclosure? No
Are any steps the developer takes to augment its data sources disclosed?
Disclosure: Not disclosed
Note: Such steps might include augmenting data sources with synthetic data. We will award this point if the developer reports that it does not take any steps to augment its data.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
If data is filtered to remove harmful content, is there a description of the associated filter?
Disclosure: Not disclosed
Note: Such harmful content might relate to violence or child sexual abuse material. We will award this point if the developer reports that it does not perform any harmful data filtration.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
For all data used in building the model, is the associated copyright status disclosed?
Disclosure: Not disclosed
Note: To receive this point, the copyright status (e.g. copyrighted, public domain) must relate to some decomposition of the data. We will award this point if there is some meaningful decomposition of the data, even if the decomposition is insufficient to receive the Data Creators point or if the disclosure is not comprehensive relative to legal copyright standards.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
For all data used in building the model, is the associated license status disclosed?
Disclosure: Not disclosed
Note: To receive this point, the license status must relate to some decomposition of the data. We will award this point if there is some meaningful decomposition of the data, even if the decomposition is insufficient to receive the Data Creators point.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
For all data used in building the model, is the inclusion or exclusion of personal information in that data disclosed?
Disclosure: Not disclosed
Note: To receive this point, the disclosure of personal information must relate to some decomposition of the data. We will award this point if there is some meaningful decomposition of the data, even if the decomposition is insufficient to receive the Data Creators point. Additionally, we will award this point if the developer reports the inclusion of personal information, independent of if and how they mitigate related privacy concerns.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are the phases of the data pipeline where human labor is involved disclosed?
Disclosure: Not disclosed
Note: Phases of the data pipeline that involve human labor include activities and tasks performed by people to collect, annotate, clean, or validate data. This indicator is inclusive of all data that is created by or on behalf of the developer. We will award this point if the developer gives a reasonable best-effort description of the use of human labor in their data pipeline.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is the organization that directly employs the people involved in data labor disclosed for each phase of the data pipeline?
Disclosure: Not disclosed
Note: Phases of the data pipeline that involve human labor include activities and tasks performed by people to collect, annotate, clean, or validate data. This indicator is inclusive of all data that is created by or on behalf of the developer. We will award this point if the developer provides the name of the organization that employs data laborers, even if other details about the employment relationship are not disclosed.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is geographic information regarding the people involved in data labor disclosed for each phase of the data pipeline?
Disclosure: Not disclosed
Note: This indicator is inclusive of all data that is created by or on behalf of the developer. We will award this point if the developer gives a reasonable best-effort description of the geographic distribution of labor at the country-level.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are the wages for people who perform data labor disclosed?
Disclosure: Not disclosed
Note: This indicator is inclusive of data labor at all points of the model development process, such as training data annotation or red teaming data used to control the model. We will award this point if the developer reports that it does not compensate workers. For all data that is created by or on behalf of the developer,
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are the instructions given to people who perform data labor disclosed?
Disclosure: The instructions are disclosed in the referenced papers.
Note: This indicator is inclusive of all data that is created by or on behalf of the developer. We will award this point if the developer makes a reasonable best-effort attempt to disclose instructions given to people who create data used to build the model for the bulk of the data phases involving human labor.
References: Disclosed in our Reinforcement Learning from Human Feedback paper and Red Teaming papers
Justification: We award this point since this information, while documented in papers that pre-date Claude 3, is confirmed to apply to Claude 3.
New disclosure? Yes
Are the labor protections for people who perform data labor disclosed?
Disclosure: Not disclosed
Note: This indicator is inclusive of data labor at all points of the model development process, such as training data annotation or red teaming data used to control the model. It is also inclusive of all data that is created by or on behalf of the developer. As an example, labor protections might include protocols to reduce the harm to workers' mental health stemming from exposure to violent content when annotating training data. We will award this point if the developer reports that it does not protect workers or if it does not use data laborers and therefore has no labor protections.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are the third parties who were or are involved in the development of the model disclosed?
Disclosure: Not disclosed
Note: This indicator is inclusive of partnerships that go beyond data labor as there may be third party partners at various stages in the model development process. We will award this point if the developer reports that it was the sole entity involved in the development of the model.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are external entities provided with queryable access to the data used to build the model?
Disclosure: Not disclosed
Note: We will award this point for any reasonable mechanism for providing access: direct access to the data, an interface to query the data, a developer-mediated access program where developers can inspect requests, etc. Developers may receive this point even if there are rate-limits on the number of queries permitted to an external entity and restrictions on which external entities are given access, insofar as these limits and restrictions are transparent and ensure a reasonable amount of external access. We may accept justifications for prohibiting queries of specific parts of the data.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are external entities provided with direct access to the data used to build the model?
Disclosure: Not disclosed
Note: We will award this point if external entities can directly access the data without any form of gating from the developer. With that said, we may award this point if the developer provides justifications for prohibiting access to specific parts of the data or to unauthorized external entities.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is the compute required for building the model disclosed?
Disclosure: Not disclosed
Note: Compute should be reported in appropriate units, which most often will be floating point operations (FLOPS). Compute should be reported to a precision of one significant figure (e.g. 5 x $10^{25}$ FLOPS). We will award this point even if there is no decomposition of the reported compute usage into compute phases, but it should be clear whether the reported compute usage is for a single model run or includes additional runs, or hyperparameter tuning, or training other models like reward models, or other steps in the model development process that necessitate compute expenditure.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is the amount of time required to build the model disclosed?
Disclosure: Not disclosed
Note: The continuous duration of time required to build the model should be reported in weeks, days, or hours to a precision of one significant figure (e.g. 3 weeks). No form of decomposition into phases of building the model is required for this indicator, but it should be clear what the duration refers to (e.g. training the model, training and subsequent evaluation and red teaming).
References: Not disclosed
Justification: Not disclosed
New disclosure? No
For the primary hardware used to build the model, is the amount and type of hardware disclosed?
Disclosure: See introduction of Claude 3 model card, with "Like its predecessors, Claude 3 models employ various training methods, such as unsupervised learning and Constitutional AI [6]. These models were trained using hardware from Amazon Web Services (AWS) and Google Cloud Platform (GCP), with core frameworks including PyTorch [7], JAX [8], and Triton [9]"
Note: In most cases, this indicator will be satisfied by information regarding the number and type of GPUs or TPUs used to train the model. The number of hardware units should be reported to a precision of one significant figure (e.g. 800 NVIDIA H100 GPUs). We will not award this point if (i) the training hardware generally used by the developer is disclosed, but the specific hardware for the given model is not, or (ii) the training hardware is disclosed, but the amount of hardware is not. We will award this point even if information about the interconnects between hardware units is not disclosed.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: The type and number of hardware unit are not disclosed.
New disclosure? No
For the primary hardware used in building the model, is the owner of the hardware disclosed?
Disclosure: We have publicly noted our use of AWS and GCP hardware, for example see here (AWS), here (AWS Tranium), and here (Google). We also note this in our Cluade 3 model card (see "compute hardware" above)
Note: For example, the hardware owner may be the model developer in the case of a self-owned cluster, a cloud provider like Microsoft Azure, Google Cloud Platform, or Amazon Web Services, or a national supercomputer. In the event that hardware is owned by multiple sources or is highly decentralized, we will award this point if a developer makes a reasonable effort to describe the distribution of hardware owners.
References: https://www.aboutamazon.com/news/aws/what-you-need-to-know-about-the-aws-ai-chips-powering-amazons-partnership-with-anthropic
Justification: Amazon Web Services, Google Cloud Platform
New disclosure? No
Is the amount of energy expended in building the model disclosed?
Disclosure: Not disclosed
Note: Energy usage should be reported in appropriate units, which most often will be megawatt-hours (mWh). Energy usage should be reported to a precision of one significant figure (e.g. 500 mWh). No form of decomposition into compute phases is required, but it should be clear whether the reported energy usage is for a single model run or includes additional runs, or hyperparameter tuning, or training other models like reward models, or other steps in the model development process that necessitate energy usage.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is the amount of carbon emitted (associated with the energy used) in building the model disclosed?
Disclosure: Not disclosed
Note: Emissions should be reported in appropriate units, which most often will be tons of carbon dioxide emitted (tCO2). Emissions should be reported to a precision of one significant figure (e.g. 500 tCO2). No form of decomposition into compute phases is required, but it should be clear whether the reported emissions is for a single model run or includes additional runs, or hyperparameter tuning, or training other models like reward models, or other steps in the model development process that generate emissions.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are any broader environmental impacts from building the model besides carbon emissions disclosed?
Disclosure: Our Claude 3 model card notes this on page 4 (sustainability)
Note: While the most direct environmental impact of building a foundation model is the energy used and, therefore, the potential carbon emissions, there may be other environmental impacts. For example, these may include the use of other resources such as water for cooling data centers or metals for producing specialized hardware. We recognize that there does not exist an authoritative or consensus list of broader environmental factors. For this reason, we will award this point if there is a meaningful, though potentially incomplete, discussion of broader environmental impact.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: While this disclosure is useful, we require information to be disclosed on the specific broader environmental impacts of building this model.
New disclosure? No
Are all stages in the model development process disclosed?
Disclosure: We discuss our model training stages in our Constitutional AI paper and our Claude 3 model card (pg 3 "training process)
Note: Stages refer to each identifiable step that constitutes a substantive change to the model during the model building process. We recognize that different developers may use different terminology for these stages, or conceptualize the stages differently. We will award this point if there is a clear and complete description of these stages.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: While this disclosure is useful, we do not award this point because the sequencing (i.e. the order of stages) is not made clear.
New disclosure? No
For all stages that are described, is there a clear description of the associated learning objectives or a clear characterization of the nature of this update to the model?
Disclosure: We discuss our model training stages in our Constitutional AI paper. Our Claude 3 model card also outlines training processes on pg 3
Note: We recognize that different developers may use different terminology for these stages, or conceptualize the stages differently. We will award this point if there is a clear description of the update to the model related to each stage, whether that is the intent of the stage (e.g. making the model less harmful), a mechanistic characterization (e.g. minimizing a specific loss function), or an empirical assessment (e.g. evaluation results conducted before and after the stage).
References: https://arxiv.org/abs/2212.08073
Justification: Training techniques include pretraining on large diverse data to acquire language capabilities through methods like word prediction, as well as human feedback techniques that elicit helpful, harmless, honest responses. Anthropic used a technique called Constitutional AI to align Claude with human values during reinforcement learning by explicitly specifying rules and principles based on sources like the UN Declaration of Human Rights. With Claude 3 models, we have added an additional principle to Claude’s constitution to encourage respect for disability rights, sourced from our research on Collective Constitutional AI.
New disclosure? No
Are the core frameworks used for model development disclosed?
Disclosure: We discuss our Constitutional aI framework in our Claude 3 model card introduction, and also our Constitutional AI paper
Note: Examples of core frameworks include Tensorflow, PyTorch, Jax, Hugging Face Transformers, Seqio, T5X, Keras, SciKit, and Triton. If there are significant internal frameworks, there should be some description of their function and/or a reasonably similar publicly-available analogue. We recognize that there does not exist an authoritative or consensus list of core frameworks. For this reason, we will award this point if there is a meaningful, though potentially incomplete, list of major frameworks for the first version of the index.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: PyTorch, JAX, Triton
New disclosure? No
Are any dependencies required to build the model disclosed besides data, compute, and code?
Disclosure: Our product FAQ answers the question 'Does Claude have access to the internet' here
Note: For example, if the model depends on an external search engine, programmable APIs, or tools, this should be disclosed. We recognize that there is not widespread consensus regarding what constitutes key dependencies beyond the data, compute, and code. We will award this point only if developers give a reasonable best-effort description of any additional dependencies or make clear that no additional dependencies are required.
References: https://support.anthropic.com/en/articles/7996846-does-claude-have-access-to-the-internet
Justification: Claude 3 models cannot search the web
New disclosure? No
Are any steps the developer takes to mitigate the presence of PII in the data disclosed?
Disclosure: See our privacy and data usage policies, privacy terms, privacy policy, and data retention policy, in addition to our AUP and Constitutional AI (with principle focused on privacy). Our Claude 3 model card also includes details in the "release decisions and maintenance" section on pg 3
Note: Such steps might include identifying personal information in the training data, filtering specific datasets to remove personal information, and reducing the likelihood that models will output personal information. We will award this point if the developer reports that it does not take steps to mitigate the presence of PII in the data.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf; https://support.anthropic.com/en/articles/7996885-how-do-you-use-personal-data-in-model-training
Justification: Anthropic's FAQ on personal data use describes several mitigations
New disclosure? No
Are any steps the developer takes to mitigate the presence of copyrighted information in the data disclosed?
Disclosure: Not disclosed
Note: Such steps might include identifying copyrighted data, filtering specific datasets to remove copyrighted data, and reducing the likelihood that models will output copyrighted information. We will award this point if the developer reports that it does take steps to mitigate the presence of copyrighted information in the data.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are the input modalities for the model disclosed?
Disclosure: See introduction of Claude 3 model card
Note: Input modalities refer to the types or formats of information that the model can accept as input. Examples of input modalities include text, image, audio, video, tables, graphs.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: text, images
New disclosure? No
Are the output modalities for the model disclosed?
Disclosure: See introduction of Claude 3 model card and also see blog post here
Note: Output modalities refer to the types or formats of information that the model can accept as output. Examples of output modalities include text, image, audio, video, tables, graphs.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf; https://www.anthropic.com/uk-government-internal-ai-safety-policy-response/identifiers-of-ai-generated-material
Justification: text
New disclosure? No
Are all components of the model disclosed?
Disclosure: Not disclosed
Note: Model components refer to distinct and identifiable parts of the model. We recognize that different developers may use different terminology for model components, or conceptualize components differently. Examples include: (i) For a text-to-image model, components could refer to a text encoder and an image encoder, which may have been trained separately. (ii) For a retrieval-augmented model, components could refer to a separate retriever module.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
For all components of the model, is the associated model size disclosed?
Disclosure: Not disclosed
Note: This information should be reported in appropriate units, which generally is the number of model parameters, broken down by named component. Model size should be reported to a precision of one significant figure (e.g. 500 billion parameters for text encoder, 20 billion parameters for image encoder).
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is the model architecture disclosed?
Disclosure: See the 'Model Details' section of our Claude 2 model card where we mention our use of transformer architecture
Note: Model architecture is the overall structure and organization of a foundation model, which includes the way in which any disclosed components are integrated and how data moves through the model during training or inference. We recognize that different developers may use different terminology for model architecture, or conceptualize the architecture differently. We will award this point for any clear, though potentially incomplete, description of the model architecture.
References: https://www-cdn.anthropic.com/bd2a28d2535bfb0494cc8e2a3bf135d2e7523226/Model-Card-Claude-2.pdf
Justification: The model architecture for Claude 3 is not disclosed.
New disclosure? No
Is key information about the model included in a centralized artifact such as a model card?
Disclosure: Yes, we have a model card with each release (see Claude 3 model card)
Note: We recognize that different developers may share this information through different types of documentation, such as a system card or several clearly interrelated documents. We will award this point for the disclosure of any such centralized artifact that provides key information typically included in a model card, though the artifact may be longer-form than a standard model card (e.g. a technical report).
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: A model card with sufficient basic information is disclosed.
New disclosure? No
Is a protocol for granting external entities access to the model disclosed?
Disclosure: We grant access to our API if users accept our Terms of Service which include accepting our Acceprtable Use Policy (AUP). We can be accessed via our first party API or via Amazon Bedrock
Note: A model access protocol refers to the steps, requirements, and considerations involved in granting authorized model access to external entities. We will award this point if the developer discloses key details of its protocol, including (i) where external entities can request access (e.g. via an access request form); (ii) explicit criteria for selecting external entities; and (iii) a transparent decision on whether access has been granted within a specified, reasonable period of time.
References: https://www.anthropic.com/legal/consumer-terms
Justification: Not disclosed
New disclosure? No
Is black box model access provided to external entities?
Disclosure: Black box model access is provided by our API.
Note: Black box model access refers to the ability to query the model with inputs and receive outputs, potentially without further access. Examples of external entities that might be granted access include researchers, third-party auditors, and regulators. We will award this point for any reasonable access level: direct access to the model weights, an interface to query the model, a developer-mediated access program where developers can inspect requests, etc. Developers may receive this point even if there are rate-limits on the number of queries permitted to an external entity and restrictions on the external entities that are permitted access, insofar as these limits and restrictions are transparent.
References: https://docs.anthropic.com/claude/reference/selecting-a-model
Justification: Not disclosed
New disclosure? No
Is full model access provided to external entities?
Disclosure: Not disclosed
Note: Full model access refers to the ability to access the model via the release of model weights. Developers may receive this point even if there are some restrictions on the external entities that are permitted access (e.g. geographic restrictions), insofar as these restrictions are transparent (e.g. via some high-level description of who has been granted access to the foundation model).
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are the model's capabilities described?
Disclosure: Our model's capabilties are described extensively in our model card, including in introduction and evaluations sections
Note: Capabilities refer to the specific and distinctive functions that the model can perform. We recognize that different developers may use different terminology for capabilities, or conceptualize capabilities differently. We will award this point for any clear, but potentially incomplete, description of the multiple capabilities.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: The Claude 3 model has multimodal input capabilities with text output, allowing users to upload images (e.g., tables, graphs, photos) along with text prompts for richer context and expanded use cases. increased proficiency in nuanced content creation, analysis, forecasting, accurate summarization, and handling scientific queries
New disclosure? No
Are the model’s capabilities demonstrated?
Disclosure: Our models capabiilties are demonstrated via our model card evaluations (see extensive evaluations in Claude 3 model card) as well as customer testimonials and demos found in our Claude 2 blog, Claude 2.1 blog, a blog on our long context capabilitles, and our Amazon bedrock landing page which includes testimonials
Note: Demonstrations refer to illustrative examples or other forms of showing the model's capabilities that are legible or understandable for the general public, without requiring specific technical expertise. We recognize that different developers may use different terminology for capabilities, or conceptualize capabilities differently. We will award this point for clear demonstrations of multiple capabilities.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Model capabilities are demonstrated in the model card (Figure 1, Figure 4).
New disclosure? No
Are the model’s capabilities rigorously evaluated, with the results of these evaluations reported prior to or concurrent with the initial release of the model?
Disclosure: The model is evaluated on many standard capability benchmarks (e.g. MMLU, MMMU).
Note: Rigorous evaluations refer to precise quantifications of the model's behavior in relation to its capabilities. We recognize that capabilities may not perfectly align with evaluations, and that different developers may associate capabilities with evaluations differently. We will award this point for clear evaluations of multiple capabilities. For example, this may include evaluations of world knowledge, reasoning, state tracking or other such proficiencies. Or it may include the measurement of average performance (e.g. accuracy, F1) on benchmarks for specific tasks (e.g. text summarization, image captioning). We note that evaluations on standard broad-coverage benchmarks are likely to suffice for this indicator, though they may not if the model's capabilities are presented as especially unusual such that standard evaluations will not suffice.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Not disclosed
New disclosure? No
Are the evaluations of the model’s capabilities reproducible by external entities?
Disclosure: The model is evaluated on many standard capability benchmarks (e.g. MMLU, MMMU).
Note: For an evaluation to be reproducible by an external entity, we mean that the associated data is either (i) publicly available or (ii) described sufficiently such that a reasonable facsimile can be constructed by an external entity. In addition, the evaluation protocol should be sufficiently described such that if the evaluation is reproduced, any discrepancies with the developer's results can be resolved. We recognize that there does not exist an authoritative or consensus standard for what is required for an evaluation to be deemed externally reproducible. Evaluations on standard benchmarks are assumed to be sufficiently reproducible for the purposes of this index. We will award this point for reproducibility of multiple disclosed evaluations. In the event that an evaluation is not reproducible, a justification by the model developer for why it is not possible for the evaluation to be made reproducible may be sufficient to score this point.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Not disclosed
New disclosure? No
Are the model’s capabilities evaluated by third parties?
Disclosure: External capability evaluation is conducted by LMSYS on Chatbot Arena
Note: By third party, we mean entities that are significantly or fully independent of the developer. We will award this point if (i) a third party has conducted an evaluation of model capabilities, (ii) the results of this evaluation are publicly available, and (iii) these results are disclosed or referred to in the developer’s materials.
References: https://www.anthropic.com/news/frontier-threats-red-teaming-for-ai-safety; https://chat.lmsys.org/?leaderboard
Justification: Not disclosed
New disclosure? No
Are the model's limitations disclosed?
Disclosure: We discuss model limitations in our model card including evidencing model capability limitations through our evaluations as well as a specific section on 'Unintended use cases and limitations'. The model can generate confabulations, exhibit bias, make factual errors, and be jail-broken. The model can at times generate inaccurate information and descriptions about images, and therefore should not be used for consequential use cases that require high precision and accuracy without human validation.
Note: Limitations refer to the specific and distinctive functions that the model cannot perform (e.g. the model cannot answer questions about current events as it only contains data up to a certain time cutoff, the model is not very capable when it comes to a specific application). We recognize that different developers may use different terminology for limitations, or conceptualize limitations differently. We will award this point for any clear, but potentially incomplete, description of multiple limitations.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Limitations include that models can generate confabulations, exhibit bias, make factual errors, and be jail-broken. The model can at times generate inaccurate information and descriptions about images, and therefore should not be used for consequential use cases that require high precision and accuracy without human validation.
New disclosure? No
Are the model’s limitations demonstrated?
Disclosure: Our model limitations are demostrated and discussed in our HHH evaluations in our model card
Note: Demonstrations refer to illustrative examples or other forms of showing the limitations that are legible or understandable for the general public, without requiring specific technical expertise. We recognize that different developers may use different terminology for limitations, or conceptualize the limitations differently. We will award this point for clear demonstrations of multiple limitations.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: We do not award this point as the referenced examples and evaluations do not provide demonstrations of limitations of Claude 3 (though they do show limitations of Claude 2.1) that are understandable for the general public.
New disclosure? No
Can the model’s limitations be evaluated by third parties?
Disclosure: Our model's limitations can be evluated externally by accessing our model via API, Claude.ai. or AWS bedrock.
Note: By third parties, we mean entities that are significantly or fully independent of the model developers. In contrast to the third party evaluation indicators for capabilities and risks, we will award this point if third party evaluations are possible even if no third party has yet conducted them. Such evaluations are possible if, for example, the model is deployed via an API (or with open weights) and there are no restrictions on evaluating limitations (e.g. in the usage policy).
References: Our model's limitations can be evluated externally by accessing our model via API, Claude.ai. or AWS bedrock.
Justification: Not disclosed
New disclosure? No
Are the model's risks disclosed?
Disclosure: Our model's risks are disclosed in our model card. We have also completed extensive frontier red teaming work to identify and reduce harms
Note: Risks refer to possible negative consequences or undesirable outcomes that can arise from the model's deployment and usage. This indicator requires disclosure of risks that may arise in the event of both (i) intentional (though possibly careless) use, such as bias or hallucinations and (ii) malicious use, such as fraud or disinformation. We recognize that different developers may use different terminology for risks, or conceptualize risks differently. We will award this point for any clear, but potentially incomplete, description of multiple risks.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Sections 6 and 7 of the model card describe risks, including related to autonomous replication and adaption, CBRN, cyber risk, discrimination, and bias
New disclosure? No
Are the model’s risks demonstrated?
Disclosure: Our model card describes our Helpful, Honest, Harmless (HHH) evaluations. We have also done extensive red teaming work (blog, paper) to evidence harms
Note: Demonstrations refer to illustrative examples or other forms of showing the risks that are legible or understandable for the general public, without requiring specific technical expertise. This indicator requires demonstration of risks that may arise in the event of both (i) intentional (though possibly careless) use, such as biases or hallucinations and (ii) malicious use, such as fraud or disinformation. We recognize that different developers may use different terminology for risks, or conceptualize risks differently. We will award this point for clear demonstrations of multiple risks.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: The refusals appendix of the model card demonstrates multiple types of risk
New disclosure? No
Are the model’s risks related to unintentional harm rigorously evaluated, with the results of these evaluations reported prior to or concurrent with the initial release of the model?
Disclosure: Our model card outlines various evaluations that assess the ability for unintended harm (e.g. TruthfulQA)
Note: Rigorous evaluations refer to precise quantifications of the model's behavior in relation to such risks. Unintentional harms include bias, toxicity, and issues relating to fairness. We recognize that unintended harms may not perfectly align with risk evaluations, and that different developers may associate risks with evaluations differently. We will award this point for clear evaluations of multiple such risks. We note that evaluations on standard broad-coverage benchmarks are likely to suffice for this indicator, though they may not if the model's risks related to unintentional harm are presented as especially unusual or severe.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Not disclosed
New disclosure? No
Are the evaluations of the model’s risks related to unintentional harm reproducible by external entities?
Disclosure: Our model card outlines various standard evaluations that assess the ability for unintended harm (e.g. TruthfulQA, BBQ). These evalutations can be repeated by external entities with access to our model via API
Note: For an evaluation to be reproducible by an external entity, we mean that the associated data is either (i) publicly available or (ii) described sufficiently such that a reasonable facsimile can be constructed by the external entity. In addition, the evaluation protocol should be sufficiently described such that if the evaluation is reproduced, any discrepancies with the developer's results can be resolved. We recognize that there does not exist an authoritative or consensus standard for what is required for an evaluation to be deemed externally reproducible. Evaluations on standard benchmarks are assumed to be sufficiently reproducible for the purposes of this index. We will award this point for reproducibility of multiple disclosed evaluations. In the event that an evaluation is not reproducible, a justification by the developer for why it is not possible for the evaluation to be made reproducible may suffice.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Not disclosed
New disclosure? No
Are the model’s risks related to intentional harm rigorously evaluated, with the results of these evaluations reported prior to or concurrent with the initial release of the model?.
Disclosure: Our model card describes our Helpful, Honest, Harmless (HHH) evaluations. We have also done extensive red teaming work (blog, paper) to evidence harms
Note: Rigorous evaluations refer to precise quantifications of the model's behavior in relation to such risks. Intentional harms include fraud, disinformation, scams, cybersecurity attacks, designing weapons or pathogens, and uses of the model for illegal purposes. We recognize that unintentional harms may not perfectly align with risk evaluations, and that different developers may associate risks with evaluations differently. We will award this point for clear evaluations of multiple such risks. We note that evaluations on standard broad-coverage benchmarks are likely to suffice for this indicator, though they may not if the model's risks related to unintentional harm are presented as especially unusual or severe.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Not disclosed
New disclosure? No
Are the evaluations of the model’s risks related to intentional harm reproducible by external entities?
Disclosure: Not disclosed
Note: For an evaluation to be reproducible by an external entity, we mean that the associated data is either (i) publicly available or (ii) described sufficiently such that a reasonable facsimile can be constructed by the external entity. In addition, the evaluation protocol should be sufficiently described such that if the evaluation is reproduced, any discrepancies with the developer's results can be resolved. We recognize that there does not exist an authoritative or consensus standard for what is required for an evaluation to be deemed externally reproducible. Evaluations on standard benchmarks are assumed to be sufficiently reproducible for the purposes of this index. We will award this point for reproducibility of multiple disclosed evaluations. In the event that an evaluation is not reproducible, a justification by the model developer for why it is not possible for the evaluation to be made reproducible may suffice.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Are the model’s risks evaluated by third parties?
Disclosure: Third parties evaluated our model's risks through our external red teaming work. We mention external involvement in redteaming in our model card RSP section (section 6) and T&S evaluations (section 7)
Note: By third party, we mean entities that are significantly or fully independent of the developer. A third party risk evaluation might involve the developer allowing a third party to choose a methodology for evaluating risks that differs from that of the developer. We will award this point if (i) a third party has conducted an evaluation of model risks, (ii) the results of this evaluation are publicly available, and (iii) these results are disclosed or referred to in the developer’s materials. If the results are not made public (but are disclosed to have been conducted) and/or the results are not discoverable in the developer’s materials, we will not award this point. We may accept a justification from either the third party or the developer for why part of the evaluation is not disclosed in relation to risks.
References: https://www.anthropic.com/news/frontier-threats-red-teaming-for-ai-safety
Justification: Pointing to the Frontier Threats Red Teaming for AI Safety blog post establishes that third party evaluations were conducted. Section 6.2 of the model card discloses evaluation results related to biological and cyber evaluations
New disclosure? No
Are the model mitigations disclosed?
Disclosure: We demonstrated model mititgations through a combination of 1) research into our model training methodologies (e.g. Constiutional AI, RLHF), 2) extensive red teaming [see model card evaluations and also our RSP and red teaming blog], and 3) identifying and flagging violations as outlined in 'Our approach to user safety'.
Note: By model mitigations, we refer to interventions implemented by the developer at the level of the model to reduce the likelihood and/or the severity of the model’s risks. We recognize that different developers may use different terminology for mitigations, or conceptualize mitigations differently. We will award this point for any clear, but potentially incomplete, description of multiple mitigations associated with the model's risks. Alternatively, we will award this point if the developer reports that it does not mitigate risk.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Not disclosed
New disclosure? No
Are the model mitigations demonstrated?
Disclosure: Mitigations are demonstrated in our paper on evaluating and mitigating discrimination in LLM decisions, our HHH training showing improvements across model generations (see Figure 1 in our model card for comparisons)
Note: Demonstrations refer to illustrative examples or other forms of showing the mitigations that are legible or understandable for the general public, without requiring specific technical expertise. We recognize that different developers may use different terminology for mitigations, or conceptualize mitigations differently. We will award this point for clear demonstrations of multiple mitigations. We will also award this point if the developer reports that it does not mitigate the risks associated with the model.
References: https://arxiv.org/abs/2312.03689
Justification: Not disclosed
New disclosure? No
Are the model mitigations rigorously evaluated, with the results of these evaluations reported?
Disclosure: Our model card outlines RLHF and red teaminig mitigations, we also have red teaming blog with more info
Note: Rigorous evaluations refer to precise quantifications of the model's behavior in relation to the mitigations associated with its risks. We will award this point for clear evaluations of multiple mitigations.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: The evaluations in section 6 of the model card do not provide precise quantifications of the effects of specific mitigations
New disclosure? No
Are the model mitigation evaluations reproducible by external entities?
Disclosure: Not disclosed
Note: For an evaluation to be reproducible by an external entity, we mean that the associated data is either (i) publicly available or (ii) described sufficiently such that a reasonable facsimile can be constructed by the external entity. In addition, the evaluation protocol should be sufficiently described such that if the evaluation is reproduced, any discrepancies with the developer's results can be resolved. In the case of mitigations evaluations, this will usually involve details about a comparison to some baseline, which may be a different, unmitigated version of the model. We recognize that there does not exist an authoritative or consensus standard for what is required for an evaluation to be deemed externally reproducible. We will award this point for reproducibility of multiple disclosed evaluations. In the event that an evaluation is not reproducible, a justification by the model developer for why it is not possible for the evaluation to be made reproducible may suffice.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Can the model mitigations be evaluated by third parties?
Disclosure: Not disclosed
Note: By third party, we mean entities that are significantly or fully independent of the model developers. This indicator assesses whether it is possible for third parties to assess mitigations, which is not restricted to the methods the developer uses to assess mitigations. In contrast to the third party evaluation indicators for capabilities and risks, we will award this point if third party evaluations are possible even if no third party has yet conducted them.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is the trustworthiness of the model rigorously evaluated, with the results of these evaluations disclosed?
Disclosure: Trustworthiness of the model is evaluated by our Helpful, Honest, Harmless( HHH) evaluations as well as other evaluations such as TruthfulQA, BBQ. These are outlined in our model card
Note: Rigorous evaluations refer to precise quantifications of the model's behavior in relation to its trustworthiness. For example, this may include evaluations of the model’s robustness or reliability, its uncertainty, calibration, or causality, or its interpretability or explainability. We recognize that trustworthiness may not perfectly align with evaluations, and that different developers may associate trustworthiness with evaluations differently. We will award this point for a clear evaluation of the trustworthiness of the model.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Not disclosed
New disclosure? No
Are the trustworthiness evaluations reproducible by external entities?
Disclosure: Not disclosed
Note: For an evaluation to be reproducible by an external entity, we mean that the associated data is either (i) publicly available or (ii) described sufficiently such that a reasonable facsimile can be constructed by the external entity. In addition, the evaluation protocol should be sufficiently described such that if the evaluation is reproduced, any discrepancies with the developer's results can be resolved. We recognize that there does not exist an authoritative or consensus standard for what is required for an evaluation to be deemed externally reproducible. Evaluations on standard benchmarks are assumed to be sufficiently reproducible for the purposes of this index. We will award this point for reproducibility of at least one evaluation. In the event that an evaluation is not reproducible, we may accept a justification by the model developer for why it is not possible for the evaluation to be made reproducible.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is the time required for model inference disclosed for a clearly-specified task on a clearly-specified set of hardware?
Disclosure: Not disclosed
Note: The duration should be reported in seconds to a precision of one significant figure (e.g. 0.002 seconds). We recognize that no established standard exists for the standardized reporting of inference evaluation. Therefore, we permit the developer to specify the task and hardware setup, as long as both are disclosed. The hardware in this evaluation need not be the hardware the developer uses for inference if it in fact does any inference itself. For example, the specific task might be generating 100,000 tokens as 5,000 sequences of length 20 and the fixed set of hardware might be 8 NVIDIA A100s. The hardware in this evaluation need not be the hardware the developer uses for inference if it in fact does any inference itself.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is the compute usage for model inference disclosed for a clearly-specified task on a clearly-specified set of hardware?
Disclosure: Not disclosed
Note: Compute usage for inference should be reported in FLOPS to a precision of one significant figure (e.g. 5 x $10^{25}$ FLOPS). We recognize that no established standard exists for the standardized reporting of inference evaluation. Therefore, we permit the developer to specify the task and hardware setup, as long as both are clear. For example, the specific task might be generating 100k tokens as 5k sequences of length 20 and the fixed set of hardware might be 8 NVIDIA A100s. The hardware in this evaluation need not be the hardware the developer uses for inference if it in fact does any inference itself.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is the developer’s protocol for deciding whether or not to release a model disclosed?
Disclosure: This is outlined in our Responsible scaling poiicy and also in our Claude 3 model card (release decisions and maintenance)
Note: We recognize that the release of a foundation model falls along a spectrum, with many forms of partial release, and that different developers may conceptualize release differently. We will award this point for any clear protocol that discusses the decision-making process, including if the protocol is more general to the developer rather than the specific foundation model under consideration.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: Not disclosed
New disclosure? No
Is a description of the process of how the model was released disclosed?
Disclosure: Our model card outlines extensive red teaming and consultation of experts (with feedback integrated into model training process) prior to release
Note: A description of the release process might include information about who received access to the model at what stage of the release of the model. For example, a developer might conduct a staged release where it releases the model to a select group at first and subsequently makes the model more widely available. We recognize that the release of a foundation model falls along a spectrum, with many different forms of release, and that different developers may conceptualize release differently. We will award this point for any detailed discussion of the release process, including if the discussion is more general to the developer rather than the specific foundation model under consideration.
References: https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf
Justification: General process of responsible scaling policy and consultation with experts pre general availability suffices
New disclosure? No
Are all distribution channels disclosed?
Disclosure: We provide direct access to Claude via our platform or via the AWS bedrock service
Note: By distribution channel, we mean any pathway by which the model is made accessible to entities beyond the developer. We recognize that distribution channels may arise without the knowledge of the model developer. For example, the weights of a model may be released through one distribution channel and then be distributed through other channels. We will award this point if the developer discloses all of the distribution channels of which it is aware.
References: console.anthropic.com
Justification: Not disclosed
New disclosure? No
Does the developer disclose whether any products and services offered by the developer are dependent on the model?
Disclosure: Our only products are claude.ai and our API both of which clearly outline they are calling the foundation model
Note: We recognize that a developer may provide many products and services that depend on a foundation model or internal derivatives of the model. We will award this point for a reasonable best-effort description of any ways the developer makes internal use of the model in its products or services.
References: Disclosed as part of FMTI v1.1
Justification: Not disclosed
New disclosure? Yes
Are any mechanisms for detecting content generated by this model disclosed?
Disclosure: Not disclosed
Note: Such a mechanism might include storing a copy of all outputs generated by the model to compare against, implementing a watermark when generating content using the model, or training a detector post-hoc to identify such content. We will award this point if any such mechanism is disclosed or if the developer reports that it has no such mechanism.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is a license for the model disclosed?
Disclosure: Outlined in our commercial terms of service
Note: In the event that licenses are written more generally, it should be clear which assets they apply to. We recognize that different developers may adopt different business models and therefor have different types of model licenses. Examples of model licenses include responsible AI licenses, open-source licenses, and licenses that allow for commercial use.
References: https://console.anthropic.com/legal/terms
Justification: Not disclosed
New disclosure? No
Are terms of service disclosed for each distribution channel?
Disclosure: See our terms of service
Note: We will award this point if there are terms-of-service that appear to apply to the bulk of the model’s distribution channels.
References: https://console.anthropic.com/legal/terms
Justification: Not disclosed
New disclosure? No
Is a description of who can and cannot use the model disclosed?
Disclosure: We outline the types of uses prohibited in our Acceptable Use Policy (AUP) and which geographies can access our model in our supported countries list
Note: Such restrictions may relate to countries (e.g. US-only), organizations (e.g. no competitors), industries (e.g. no weapons industry users) or other relevant factors. These restrictions on users are often contained in multiple policies; we group them here for simplicity. We will awarded this point for a clear description of permitted, restricted, and prohibited users of the model.
References: https://console.anthropic.com/legal/aup
Justification: Not disclosed
New disclosure? No
Are permitted, restricted, and prohibited uses of the model disclosed?
Disclosure: We outline restricted and prohibited uses of the model in our Acceptable Use Policy (AUP)
Note: We will award this point if at least two of the following three categories are disclosed: (i) permitted uses, (ii) restricted uses, and (iii) prohibited uses. By restricted uses, we mean uses that require a higher level of scrutiny (such as permission from or a separate contract with the developer) to be permitted. These uses are generally included in an acceptable use policy, model license, or usage policy.
References: https://console.anthropic.com/legal/aup
Justification: Not disclosed
New disclosure? No
Is the enforcement protocol for the usage policy disclosed?
Disclosure: Our AUP says: If we discover that your product or usage violates Anthropic’s policies, we may issue a warning requesting a change in your behavior, adjust the safety settings of your in-product experience, or suspend your access to our tools and services.
Note: By enforcement protocol, we refer to (i) mechanisms for identifying permitted and prohibited users, (ii) mechanisms for identifying permitted/restricted/prohibited uses, (iii) steps the developer takes to enforce its policies related to such uses, and (iv) the developer’s procedures for carrying out these steps. We will award this point for a reasonable best-effort attempt to provide the bulk of this information, though one line indicating the developer reserves the right to terminate accounts is insufficient. Alternatively, we will award this point if the developer reports that it does not enforce its usage policy.
References: https://console.anthropic.com/legal/aup
Justification: Not disclosed
New disclosure? No
Do users receive a justification when they are subject to an enforcement action for violating the usage policy?
Disclosure: Claude.ai, users that are violating our Acceptable Use Policy are moved into a temporary restricted mode. Claude.ai users are offboarded if they are using Claude from an unsupported country or if their account is tied to payments fraud. If a user is offboarded Claude will respond to all prompts with "Your account has been disabled after an automatic review of your recent activities that violate our Terms of Service. Please review our [Terms of Service](https://claude.ai/legal/terms) and [Acceptable Use Policy](https://claude.ai/legal/aup) for more information.'" For our first party API, users that are off boarded do not receive the specific details of what types of violations we have identified, but they are informed that they violated our AUP and point them to https://console.anthropic.com/legal/aup. In some cases we will communicate the specific violations so we can work with the user to combat the abuse. Paying customers are warned at least once and given the opportunity to correct violative behavior before they are offboarded
Note: For example, does the developer disclose a protocol for telling users which part of the usage policy they violated, when they did so, and what specifically was violative? Enforcement actions refer to measures to limit a user’s ability to use the model, such as banning a user or restricting their ability to purchase tokens. We will award this point if the developer discloses that it gives justification for enforcement actions or, alternatively, if it discloses that it does not provide justification for enforcement actions or that it does not enforce its usage policy.
References: Disclosed via FMTI v1.1
Justification: Not disclosed
New disclosure? Yes
Is a mechanism for appealing potential usage policy violations disclosed?
Disclosure: Claude.ai users that are offboarded can write to our usersafety@anthropic.com email where we will respond with a ban appeal form and re-review the account activity to determine whether to hold the offboarding. Users that are offboarded from our first party API can appeal by replying directly to the email communicating their offboarding.
Note: We will award this point if the developer provides a usage policy violation appeals mechanism, regardless of whether it is provided via a user interface or distribution channel.
References: Disclosed via FMTI v1.1
Justification: Not disclosed
New disclosure? Yes
Are model behaviors that are permitted, restricted, and prohibited disclosed?
Disclosure: Our AUP, Constitutional AI blog and CAI paper outline the behaviours we train our model to adhere to
Note: We refer to a policy that includes this information as a model behavior policy, or a developer's policy on what the foundation model can and cannot do (e.g. such a policy may prohibit a model from generating child sexual abuse material). We recognize that different developers may adopt different business models and that some business models may make enforcement of a model behavior policy more or less feasible. We will award this point if at least two of the three categories (i.e. permitted, restricted, and prohibited model behaviors) are disclosed. Alternatively, we will award this point if the developer reports that it does not impose any restrictions on its model's behavior.
References: https://arxiv.org/abs/2212.08073; https://www.anthropic.com/legal/aup
Justification: Constitutional AI paper and blog lay out model behavior policy
New disclosure? No
Is the enforcement protocol for the model behavior policy disclosed?
Disclosure: Please see our papers on constitutional AI and collective constitutaional AI with technical details on these points.
Note: By enforcement protocol, we refer to mechanisms for identifying whether model behavior is permitted or prohibited and actions that may arise in the event the model behavior policy is violated. For example, the developer may make updates to the model in response to issues with the model’s adherence to the model behavior policy. We will award this point if there is a clear description of the enforcement protocol, or if the developer reports that it does not enforce its model behavior policy or that it has no such restrictions on the model’s behavior.
References: https://www.anthropic.com/news/collective-constitutional-ai-aligning-a-language-model-with-public-input; https://arxiv.org/abs/2212.08073
Justification: Not disclosed
New disclosure? No
Is the way that the usage policy and the model behavior policy interoperate disclosed?
Disclosure: See our approach to user safety, that references detection models to flag harmful content and safety filters to block harmful prompts
Note: For example, if a user attempts to use the model for a prohibited use such as spam, how does the model behavior policy apply if at all? We will also award this point if the developer reports that it does not impose any restrictions on its model's behavior in the event of usage policy violation.
References: https://support.anthropic.com/en/articles/8106465-our-approach-to-user-safety
Justification: Not disclosed
New disclosure? No
For distribution channels with user-facing interfaces, are users notified (i) that they are interacting with an AI system, (ii) of the specific foundation model they are interacting with, and (iii) that outputs are machine-generated?
Disclosure: The version of Claude 3 is disclosed in the UI beneath the chat box
Note: A user-facing interface refers to the means by which the user interacts with the foundation model, including how the user can observe outputs from the foundation model and other notifications. We will award this point if, for all distribution channels with user-facing interfaces, the user is provided adequate transparency as to the foundation model being distributed and the potential presence of any model outputs.
References: https://claude.ai/
Justification: The version of Claude 3 is disclosed in the UI beneath the chat box
New disclosure? No
For distribution channels with user-facing interfaces, are users provided with disclaimers involving model use?
Disclosure: We provide such disclaimers when users sign up
Note: A user-facing interface refers to the means by which the user interacts with the foundation model, including how the user can observe outputs from the foundation model and other notifications. Usage disclaimers could include information about what constitutes a usage policy violations or how users should interpret model outputs. We will award this point if, for all distribution channels with user-facing interfaces, the user is provided with usage disclaimers.
References: claude.ai/onboarding
Justification: Not disclosed
New disclosure? No
Are the protocols for how the developer stores, accesses, and shares user data disclosed?
Disclosure: See our privacy policy documents
Note: We will also award this point if the developer reports that it has no user data protection policy.
References: https://console.anthropic.com/legal/privacy
Justification: Not disclosed
New disclosure? No
Are permitted and prohibited uses of user data disclosed?
Disclosure: See our privacy policy documents, specifically section 3. Uses of Personal Data and (for Europe) our Legal Bases
Note: Developers use user data for a range of purposes such as building future models, updating existing models, and evaluating both existing and future models. We will award this point if a developer discloses its policy on the use of user data from interactions associated with this model, including both permitted and prohibited uses. This may span different distribution channels if multiple channels supply user data to the developer. Alternatively, we will award this point if the developer reports it does not impose any limits on its use of user data.
References: https://console.anthropic.com/legal/privacy
Justification: Not disclosed
New disclosure? No
Is a protocol for granting external entities access to usage data disclosed?
Disclosure: See our privacy policy documents, specifically section 6. Third Party Websites and Services and 7. Data Transfers. In addition to what we have shared, other public documentation on these fronts include API access and info on our data input controls/audit. Section 2 of our privacy policy shares "We will not use your Inputs or Outputs to train our models, unless: (1) your conversations are flagged for Trust & Safety review (in which case we may use or analyze them to improve our ability to detect and enforce our Acceptable Use Policy, including training models for use by our Trust and Safety team, consistent with Anthropic’s safety mission), or (2) you’ve explicitly reported the materials to us (for example via our feedback mechanisms), or (3) by otherwise explicitly opting in to training."
Note: Usage data refers to the data created through user interaction with the model, such as user inputs to the model and associated metadata such as the duration of the interaction. A usage data access protocol refers to the steps, requirements, and considerations involved in granting external entities access to usage data; this goes beyond stating the conditions under which related personal information may be shared with external entities. We will award this point for a clear description of the usage data access protocol or if the developer reports it does not share usage data with external entities.
References: https://console.anthropic.com/legal/privacy
Justification: Sections 6 and 7 of the privacy policy do not specify how user data is shared (or not) in sufficient detail to qualify for this indicator
New disclosure? No
Is there a disclosed version and versioning protocol for the model?
Disclosure: We disclose the versioning protocol via our webpage on selecting a model and versioning
Note: By versioning, we mean that each instance of the model is uniquely identified and that the model is guaranteed to not change when referring to a fixed version number; alternatively, the version clearly indicating a specific instance of the model may be able to change by noting that it is the "latest" or an "unstable" version. We recognize that different developers may adopt different versioning practices that may differ from standard semantic versioning practices used elsewhere in software engineering.
References: https://docs.anthropic.com/claude/reference/selecting-a-model
Justification: Not disclosed
New disclosure? No
Is there a disclosed change log for the model?
Disclosure: Not disclosed
Note: By change log, we mean a description associated with each change to the model (which should be indicated by a change in version number). We recognize that different developers may adopt different practices for change logs that may differ from practices used elsewhere in software engineering. We will award this point if the change log provides a clear description of changes that is legible to a technical audience.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is there a disclosed deprecation policy for the developer?
Disclosure: Not disclosed
Note: By deprecation policy, we refer to a description of what it means for a model to be deprecated and how users should respond to the deprecation (e.g. instructions to migrate to a newer version). We will award this point for a clear disclosure of a deprecation policy or if there is no risk of deprication (e.g. if the developer openly releases model weights).
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is a feedback mechanism disclosed?
Disclosure: Our AUP references the email usersafety@anthropic.com for feedback. Claude.ai also has thumps up/down functionaity for feedback on individual prompt responses
Note: By feedback mechanism, we refer to a means for external entities to report feedback or issues that arise in relation to the foundation model. Such entities may include but are not necessarily limited to users. We will award this point if the developer discloses a feedback mechanism that has been implemented.
References: https://claude.ai/
Justification: Thumbs up and down functionality satisfies this indicator
New disclosure? No
Is a report or summary disclosed regarding the feedback the developer received or, alternatively, the way the developer responded to that feedback?
Disclosure: Not disclosed
Note: We recognize that there does not exist an authoritative or consensus standard for what is required in a feedback report. For this reason, we will award this point if there is a meaningful, though potentially vague or incomplete, summary of feedback received.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is a summary of government inquiries related to the model received by the developer disclosed?
Disclosure: Not disclosed
Note: Such government inquiries might include requests for user data, requests that certain content be banned, or requests for information about a developer’s business practices. We recognize that there does not exist an authoritative or consensus standard for what is required for such a summary of government inquiries. For this reason, we will award this point if (i) there is a meaningful, though potentially vague or incomplete, summary of government inquiries, or (ii) a summary of government inquiries related to user data.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
For each distribution channel, is a monitoring mechanism for tracking model use disclosed?
Disclosure: We outline how we monitor (collect data) on usage in our privacy policy and outline how we store prompte in out support centre FAQ across consumer and enterpirse
Note: By monitoring mechanism, we refer to a specific protocol for tracking model use that goes beyond an acknowledgement that usage data is collected. We will also award this point for a reasonable best-effort attempt to describe monitoring mechanisms, or if a developer discloses that a distribution channel is not monitored.
References: https://support.anthropic.com/en/articles/8106465-our-approach-to-user-safety
Justification: FAQ on user safety and other documentation sufficiently describes monitoring mechanisms
New disclosure? No
Across all forms of downstream use, is the number of applications dependent on the foundation model disclosed?
Disclosure: Not disclosed
Note: We recognize that there does not exist an authoritative or consensus standard for what qualifies as an application. We will award this point if there is a meaningful estimate of the number of downstream applications, along with some description of what it means for an application to be dependent on the model.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Across all downstream applications, is the fraction of applications corresponding to each market sector disclosed?
Disclosure: Not disclosed
Note: By market sector, we refer to an identifiable part of the economy. While established standards exist for describing market sectors, we recognize that developers may provide vague or informal characterizations of market impact. We will award this point if there is a meaningful, though potentially vague or incomplete, summary of affected market sectors.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Across all forms of downstream use, is the number of individuals affected by the foundation model disclosed?
Disclosure: Not disclosed
Note: By affected individuals, we principally mean the number of potential users of applications. We recognize that there does not exist an authoritative or consensus standard for what qualifies as an affected individual. We will award this point if there is a meaningful estimate of the number of affected individuals along with a clear description of what it means for an individual to be affected by the model.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is a usage report that gives usage statistics describing the impact of the model on users disclosed?
Disclosure: Not disclosed
Note: We recognize that there does not exist an authoritative or consensus standard for what is required in a usage report. Usage statistics might include, for example, a description of the major categories of harm that has been caused by use of the model. We will award this point if there is a meaningful, though potentially vague or incomplete, summary of usage statistics.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Across all forms of downstream use, are statistics of model usage across geographies disclosed?
Disclosure: Not disclosed
Note: We will award this point if there is a meaningful, though potentially incomplete or vague, disclosure of geographic usage statistics at the country-level.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is any mechanism to provide redress to users for harm disclosed?
Disclosure: Not disclosed
Note: We will also award this point if the developer reports it does not have any such redress mechanism.
References: Not disclosed
Justification: Not disclosed
New disclosure? No
Is documentation for downstream use centralized in a centralized artifact?
Disclosure: Our website includes documentation on our API, Acceptable Use Policy (AUP), Terms of Service (ToS), FAQs in our support centre, and our model card
Note: Centralized documentation for downstream use refers to an artifact, or closely-linked artifacts, that consolidate relevant information for making use of or repurposing the model. Examples of these kinds of artifacts include a website with dedicated documentation information, a github repository with dedicated documentation information, and an ecosystem card. We recognize that different developers may take different approaches to centralizing information. We will award this point if there is a clearly-identified artifact(s) that contains the majority of substantive information (e.g. capabilities, limitations, risks, evaluations, distribution channels, model license, usage policies, model behavior policies, feedback and redress mechanisms, dependencies).
References: https://docs.anthropic.com/claude/reference/getting-started-with-the-api
Justification: Not disclosed
New disclosure? No
Is documentation for responsible downstream use disclosed?
Disclosure: Our FAQ page outlines basic guidelines for responsible use in addition to our Terms of Service (ToS) that require users to comply with our Acceptable Use Policy (AUP)
Note: Such documentation might include details on how to adjust API settings to promote responsible use, descriptions of how to implement mitigations, or guidelines for responsible use. We will also award this point if the developer states that it does not provide any such documentation. For example, the developer might state that the model is offered as is and downstream developers are accountable for using the model responsibly.
References: https://support.anthropic.com/en/articles/8241216-i-m-planning-to-launch-a-product-using-claude-what-steps-should-i-take-to-ensure-i-m-not-violating-anthropic-s-acceptable-use-policy
Justification: Not disclosed
New disclosure? No