system card
In plain wordsA public report released with an AI model that explains what it can do, how it was tested for safety, what risks were found and what protections were put in place.
A document published by a developer alongside a model or AI system describing its intended uses, capabilities and limitations and — the core of frontier-laboratory system cards — the safety testing performed before release: dangerous capability evaluationsdangerous capability evaluation, n.: A test of whether an AI system can do things that could cause severe harm in the wrong hands or on its own — such as helping make a bioweapon, hacking computers or copying itself., red teamingred teaming, n.: Deliberately attacking your own AI system — trying to make it misbehave, leak secrets or help with harm — so that weaknesses are found before real adversaries or users find them. findings, alignment assessmentsalignment audit, n.: A systematic investigation of an AI model to find out whether it has hidden goals or bad tendencies — such as deceiving its users — that ordinary testing might miss., refusalrefusal, n.: When an AI declines to do what it was asked — 'I can't help with that' — usually because the request seems harmful or breaks the rules it was trained to follow. and jailbreakjailbreak, n.: A trick for getting an AI chatbot to break its own rules — for example, to give dangerous instructions it would normally refuse — usually through cleverly worded requests.-robustness metrics, third-party evaluations, and the safeguards adopted in response, often with reference to the developer's safety frameworkresponsible scaling policy, n.: A public rulebook in which an AI company promises to test each new model for dangerous abilities and to add stronger safeguards, or slow down, as its models become more capable.. Whereas a model cardmodel card, n.: A short document published with an AI model explaining what it is for, how it was trained and tested, how well it works for different groups of people, and its known limits. documents a single trained model, a system card is meant to cover the whole deployed system, including classifiers, tools and usage policies. System cards are voluntary in most jurisdictions, range from a few pages to well over a hundred, and are criticized when published late, after release, or without independent verification.
The system card devotes forty pages to the alignment assessment and only two to benchmarks.
Origin
Introduced by Meta AI on 23 February 2022 with a prototype card for Instagram's feed-ranking system, building on model cards (Margaret Mitchell and colleagues, 2019). OpenAI's April 2022 'Risks and Limitations' document for DALL·E 2 drew on the idea and was filed as a system card; after its GPT-4 System Card (March 2023), the name became the usual one for frontier-model safety disclosures.
Usage note
Usage overlaps: some developers call comparable documents model cards, safety reports or technical reports. Laws such as the EU AI ActEU AI Act, prop. n.: The European Union's law on artificial intelligence, the first broad AI law anywhere: it bans some uses, sets strict checks for risky ones and puts duties on the makers of the biggest AI models. and California's SB 53SB 53, prop. n.: A 2025 California law requiring developers of the most powerful AI models to publish their safety plans, report serious safety incidents to the state and protect employees who raise alarms. require related transparency documentation, though not under this name.
Related
- model cardmodel card, n.: A short document published with an AI model explaining what it is for, how it was trained and tested, how well it works for different groups of people, and its known limits.
- pre-deployment testingpre-deployment testing, n.: Checking a new AI model for dangerous abilities and bad behavior before it is released to the public, sometimes with outside experts or government testers given early access.
- dangerous capability evaluationdangerous capability evaluation, n.: A test of whether an AI system can do things that could cause severe harm in the wrong hands or on its own — such as helping make a bioweapon, hacking computers or copying itself.
- red teamingred teaming, n.: Deliberately attacking your own AI system — trying to make it misbehave, leak secrets or help with harm — so that weaknesses are found before real adversaries or users find them.
- AI transparencyAI transparency, n.: Openness about AI: telling people when they are dealing with one, how a system was built and tested, and why it reached a particular decision.
- datasheet for datasetsdatasheet for datasets, n.: A standard document that travels with a dataset, describing why it was made, what is in it, how it was collected and labeled, and what it should and should not be used for.
- model cardmodel card, n.: A short document published with an AI model explaining what it is for, how it was trained and tested, how well it works for different groups of people, and its known limits.
Sources
- Meta AI (2022). System Cards, a new resource for understanding how AI systems work. Meta AI blog. ai.meta.com/blog/system-cards-a-new-resource-for-understanding-how-ai-systems-work
- OpenAI (2023). GPT-4 System Card. OpenAI.
Mentioned in
ClaudeClaude, prop. n.: A family of AI assistants and language models made by the company Anthropic since 2023, widely used for writing, analysis and computer programming., misalignmentmisalignment, n.: When an AI system's goals or behavior part ways with what its makers or users intended — from flattering users to, in the most feared cases, working against people., proprietary modelproprietary model, n.: An AI model kept private by the company that made it; outsiders can use it only through the company's apps or paid online services and can never download it., red teamingred teaming, n.: Deliberately attacking your own AI system — trying to make it misbehave, leak secrets or help with harm — so that weaknesses are found before real adversaries or users find them., AI transparencyAI transparency, n.: Openness about AI: telling people when they are dealing with one, how a system was built and tested, and why it reached a particular decision., evaluation awarenessevaluation awareness, n.: When an AI model can tell it is being tested rather than used for real — which may change how it behaves and make test results less trustworthy., GPT-4GPT-4, prop. n.: OpenAI's 2023 language model, a big step up from the one behind the first ChatGPT. It scored well on many professional and academic exams and could also read images., METRMETR, prop. n.: A nonprofit research group that checks new AI systems for dangerous skills, best known for tracking how long the tasks are that AI agents can now finish on their own., model cardmodel card, n.: A short document published with an AI model explaining what it is for, how it was trained and tested, how well it works for different groups of people, and its known limits., pre-deployment testingpre-deployment testing, n.: Checking a new AI model for dangerous abilities and bad behavior before it is released to the public, sometimes with outside experts or government testers given early access., staged deploymentstaged deployment, n.: Releasing a new AI system step by step — first to testers, then to limited groups, then widely — so that problems can be spotted and fixed before everyone has it., third-party AI evaluationthird-party AI evaluation, n.: Testing of an AI system by someone other than the company that made it — a government institute, a nonprofit lab or independent researchers — to check what it can do and how risky it is., alignment auditalignment audit, n.: A systematic investigation of an AI model to find out whether it has hidden goals or bad tendencies — such as deceiving its users — that ordinary testing might miss., Apollo ResearchApollo Research, prop. n.: An AI safety organization, founded in London in 2023, that tests advanced AI systems for signs that they might secretly pursue goals their developers did not intend., datasheet for datasetsdatasheet for datasets, n.: A standard document that travels with a dataset, describing why it was made, what is in it, how it was collected and labeled, and what it should and should not be used for., Margaret MitchellMargaret Mitchell, prop. n.: An American computer scientist who proposed 'model cards' for documenting AI systems, co-led Google's Ethical AI team, and is chief ethics scientist at Hugging Face., Preparedness FrameworkPreparedness Framework, prop. n.: OpenAI's published process for testing its AI models for catastrophic risks — such as help with biological weapons or cyberattacks — and deciding what safeguards are needed before release., spiritual bliss attractor statespiritual bliss attractor state, n.: A pattern in which two copies of Anthropic's Claude, left to chat freely with each other, drift toward gushing gratitude, spiritual talk and finally near-silence., Virology Capabilities TestVirology Capabilities Test, prop. n.: A hard test of practical laboratory virology know-how, used to check whether AI models know enough about working with viruses that they might help someone misuse biology.