Lesson 243 of 2116
UK AI Safety Institute
The UK stood up the world's first government AI safety institute in November 2023. Its structure, scope, and access model are templates other nations are following.
Lesson map
What this lesson covers
Learning path
The main moves in order
- 1From Bletchley to Institute
- 2AISI
- 3pre-deployment testing
- 4Bletchley
Concept cluster
Terms to connect while reading
Section 1
From Bletchley to Institute
The UK hosted the first global AI Safety Summit at Bletchley Park in November 2023. Coming out of it, Prime Minister Rishi Sunak announced the AI Safety Institute (AISI), a government body tasked with evaluating frontier AI. It began operating immediately with a direct mandate from the Department for Science, Innovation and Technology.
What AISI does
- Pre-deployment evaluations of frontier models from OpenAI, Anthropic, Google DeepMind, Meta, and others
- Research on misuse risks — cyber, bio, autonomous systems, deepfakes
- Public publication of evaluation results and methodologies
- Coordination with the US AI Safety Institute (later CAISI) and similar bodies
- Grants to external researchers and open-source safety tools
Staffing and output
- 1Employs dozens of researchers including alumni from DeepMind, OpenAI, Anthropic
- 2Led initially by Ian Hogarth and Jade Leung, with Geoffrey Irving as research director
- 3Published detailed evaluation reports on Claude, GPT-4o, o1, Llama, and others throughout 2024-2025
- 4Identifies failure modes that inform labs' own safety work
“We are not a regulator. We are a source of evidence. What governments do with that evidence is up to them.”
Key terms in this lesson
The big idea: AISI showed that a government can build technical AI evaluation capacity without new laws. Every serious AI policy actor now has something like it or is building one.
End-of-lesson quiz
Check what stuck
15 questions · Score saves to your progress.
Tutor
Curious about “UK AI Safety Institute”?
Ask anything about this lesson. I’ll answer using just what you’re reading — short, friendly, grounded.
Progress saved locally in this browser. Sign in to sync across devices.
Related lessons
Keep going
Creators · 45 min
Red-Teaming: The Ethics of Breaking AI on Purpose
Red-teamers get paid to make AI misbehave. The field has grown into a real discipline — with its own methods, its own ethics, and its own unresolved questions.
Creators · 45 min
The EU AI Act: The Global Floor, Whether You Like It or Not
The EU AI Act is the most sweeping AI law in the world. It will set the compliance floor for anyone who ships globally. Here is the architecture, the timeline, and what it gets right and wrong.
Creators · 40 min
Jailbreak Case Studies: What Actually Broke
Abstract jailbreak theory is less useful than real cases. Here are the techniques that worked on production models, what they taught us, and what is still unsolved.