A public report an AI company releases with a model, describing what it can do, how it was tested for risks and what safeguards it has.
AI Drafted with AI help · Reviewed by Keshav Aggarwal · Updated
In simple words
A new medicine comes with a leaflet that says what it is for, what tests it passed and what side effects to watch for.
A system card is that leaflet for an AI model: it tells you its strengths, its risks and how the company tried to keep it safe.
The technical side
A system card, also called a model card, is a document published alongside an AI model that reports its intended uses, test results, safety evaluations, red-teaming findings and safeguards. The idea goes back to the 2019 paper “Model Cards for Model Reporting” by Margaret Mitchell and colleagues. Large labs now publish system cards for major releases, often grading risks such as cybersecurity, biological threats and self-improvement under their own safety frameworks.
On 17 November 2026 OpenAI released GPT-6.2 Astra and GPT-6.2 Sol. Their system card, under OpenAI’s Preparedness Framework, rates both models Critical for cybersecurity and High for biological and chemical risk, and rates Astra High for AI self-improvement, the first OpenAI model released with that rating, according to Unite.AI. Compare models on AI models.
System Card kya hai?
System card ek report hai jo AI company naye model ke saath jaari karti hai: model kya kar sakta hai, kaise test hua, kya khatre hain aur kaun se safety upay lage hain. 17 Nov 2026 ko OpenAI ke GPT-6.2 Astra ke system card ne cybersecurity khatre ko "Critical" bataya.
Related terms
- Red TeamingAITesting an AI by attacking it on purpose, trying to make it break rules or cause harm, so the problems are fixed before the public uses it.
- Frontier ModelAIOne of the most powerful AI models of its time, trained with huge computing power, which laws often treat as needing extra safety checks.
- AI BenchmarkAIA fixed set of tasks used to test and score AI models, so different models can be compared on the same work.
- LLM (Large Language Model)AIAn AI model trained on huge amounts of text so it can read and write language, like the models behind ChatGPT, Claude and Gemini.