model specification
In plain wordsA public document in which an AI company spells out how its models are supposed to behave — what they should do, refuse and prioritize, and whom to obey when instructions conflict.
A written statement of the behavior a developer intends its AI models to exhibit, serving both as a target for training and evaluation and as public documentation. Following the pattern of OpenAI's Model Spec (2024), such a document typically sets out overall objectives; hard rules that no instruction can override; default behaviors that deploying businesses or users may adjust; and a chain of command for resolving conflicts between instructions from the developer, the deployer and the user (see instruction hierarchyinstruction hierarchy, n.: A rule built into AI assistants about whose instructions win when they conflict: the developer's beat the user's, and the user's beat any instructions hidden in a web page or document.). It is more detailed than a usage policy, which constrains users, and more operational than a list of ethical principles. Specifications feed training directly — as data for AI-feedbackreinforcement learning from AI feedback, n.: Training an AI with preference judgments made by another AI instead of by people, which is cheaper and faster but only as good as the AI doing the judging. methods, as grading rubrics, or as the text a model learns to reason over in deliberative alignmentdeliberative alignment, n.: A training method in which an AI is taught the actual text of its safety rules and learns to think through which rules apply before it answers a sensitive request. — and make it possible to tell whether an undesirable output was a training failure or intended policy.
Under the model spec, a developer's system prompt can turn off that default but not the hard rule.
Origin
OpenAI published its Model Spec in May 2024 and released a substantially revised version into the public domain in February 2025, revising it several times since (most recently in August 2026, as of October 2026); Anthropic's constitutionAI constitution, n.: A written set of principles that an AI company uses to shape an AI's values and behavior, both during training and as a public statement of what the AI should be like. for Claude plays a similar role. 'Spec' echoes software engineering, where a specification states required behavior independently of implementation.
Usage note
A specification describes intended behavior; deployed models may diverge from it. Some are published in full and invite public comment; others remain internal.
Related
- AI constitutionAI constitution, n.: A written set of principles that an AI company uses to shape an AI's values and behavior, both during training and as a public statement of what the AI should be like.
- instruction hierarchyinstruction hierarchy, n.: A rule built into AI assistants about whose instructions win when they conflict: the developer's beat the user's, and the user's beat any instructions hidden in a web page or document.
- deliberative alignmentdeliberative alignment, n.: A training method in which an AI is taught the actual text of its safety rules and learns to think through which rules apply before it answers a sensitive request.
- character trainingcharacter training, n.: Training an AI to have a consistent, deliberate personality — traits such as curiosity, honesty and warmth — rather than leaving its manner and values as accidental by-products of other training.
- system promptsystem prompt, n.: Background instructions given to a chatbot before the conversation starts, setting its role, rules, tone and tools. Users usually never see it, but it shapes every reply.
- 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.
Mentioned in
AI alignmentAI alignment, n.: Making an AI system genuinely try to do what the people it serves want — not a literal-minded, distorted or self-serving version of it. Also the research field working on this., 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., AI constitutionAI constitution, n.: A written set of principles that an AI company uses to shape an AI's values and behavior, both during training and as a public statement of what the AI should be like., Constitutional AIConstitutional AI, n.: Anthropic's technique for making an AI less harmful by having it check and rewrite its own answers against a short written set of principles, so that people need not label harmful examples by hand., corrigibilitycorrigibility, n.: An AI's willingness to let people correct it — to accept being paused, retrained, overruled or shut down — rather than resisting or quietly working around them., helpful, honest and harmlesshelpful, honest and harmless, phr.: A three-part goal for AI assistants, popularized by Anthropic: be genuinely useful, tell the truth, and avoid causing harm — shorthand for what an aligned chatbot should be., machine ethicsmachine ethics, n.: The effort to build AI systems that can tell right from wrong and act ethically themselves, rather than only being used ethically by people., over-refusalover-refusal, n.: When an AI wrongly refuses a harmless request because it superficially resembles a harmful one — like declining to explain how to 'kill' a frozen computer program., robustnessrobustness, n.: An AI system's ability to keep working correctly and safely when conditions change — unfamiliar data, noisy inputs, or people deliberately trying to trick it., safety trainingsafety training, n.: The part of an AI's training that teaches it to avoid harmful behavior — declining dangerous requests, resisting manipulation and following its developer's rules — on top of being generally helpful., steerabilitysteerability, n.: How easily an AI model's behavior — its tone, style, persona or rules — can be adjusted by the people using or deploying it, for example through instructions, without retraining., system promptsystem prompt, n.: Background instructions given to a chatbot before the conversation starts, setting its role, rules, tone and tools. Users usually never see it, but it shapes every reply., usage policyusage policy, n.: A company's published rules on what people may and may not use its AI for — for example, no help building weapons, no scams and no sexual content involving children., AI-enabled coupAI-enabled coup, n.: A seizure of power by a small group that uses advanced AI to bypass the soldiers, officials and workers whose cooperation power grabs normally need., character trainingcharacter training, n.: Training an AI to have a consistent, deliberate personality — traits such as curiosity, honesty and warmth — rather than leaving its manner and values as accidental by-products of other training., deliberative alignmentdeliberative alignment, n.: A training method in which an AI is taught the actual text of its safety rules and learns to think through which rules apply before it answers a sensitive request., instruction hierarchyinstruction hierarchy, n.: A rule built into AI assistants about whose instructions win when they conflict: the developer's beat the user's, and the user's beat any instructions hidden in a web page or document., safe completionsafe completion, n.: A way of training AI to handle sensitive requests by giving the most helpful answer that stays safe, instead of simply choosing between answering fully and refusing.