For Heads-of · Practitioner
Safety Case
A structured argument, supported by evidence, that a system is acceptably safe for a given use in a given context. Long established in nuclear, aviation and defence engineering, safety cases are being adopted for frontier AI as a way to make safety claims explicit and checkable.
- safety case
- assurance
- claims arguments evidence
- frontier AI
- AI Security Institute
How it is built
A safety case starts from a top-level claim, such as that a model cannot meaningfully help an attacker carry out a cyber attack, and breaks it into sub-claims, each supported by an argument and evidence, typically in a claims, arguments and evidence structure. The value is that every assumption and gap becomes visible to a reviewer rather than being implied in a test report.
In AI
The UK AI Security Institute has published a template for one kind of argument, an inability argument, showing how a developer could argue from evaluation results that a model lacks a dangerous capability. Researchers are also developing arguments that rest on controls rather than inability, for models that do have dangerous capabilities, and frontier developers increasingly describe their release decisions in these terms.
For a deploying organisation
The same discipline works at any scale. For a high-stakes AI use, writing down the claim (this system is safe enough for this purpose), the argument and the evidence exposes untested assumptions before a regulator or an incident does. It pairs naturally with AI assurance, and the register entry for the assumptions it exposes is untested AI failure scenarios.
Sources
Related Terms
AI Assurance
The work of measuring, evaluating and communicating whether an AI system, or the organisation running it, meets the claims made for it, so that others can rely on those claims. It spans internal testing through to independent audit and certification, and the question it answers is not whether a system is good but whether a statement about it can be trusted.
AI Evaluation
Systematic measurement of what an AI model or system can do, how it tends to behave and how well its safeguards hold, using benchmarks, task-based tests, red teaming or studies with people. Evaluations are the evidence most AI claims rest on, and they have known weaknesses.
AI Safety
The field of research and practice concerned with making AI systems behave as intended, avoid unintended harm and stay under meaningful human control. In 2025 the UK and US government institutes renamed themselves around security, which narrowed the official focus without removing the questions an organisation deploying AI has to answer.
Frontier Safety Framework
A frontier AI developer's published commitment to test its models for dangerous capabilities and to apply stronger safeguards, or pause, when defined capability thresholds are crossed. Examples include Anthropic's Responsible Scaling Policy, OpenAI's Preparedness Framework and Google DeepMind's Frontier Safety Framework.