Government & Politics

UK AI Security Institute (AISI)

High46% confidenceProfile verified Jul 23, 2026

The UK government's frontier-AI evaluation body, rebranded from the AI Safety Institute to the AI Security Institute in 2025 to emphasize protecting citizens, institutions and democratic values from AI misuse. The Institute was given access by Anthropic to evaluate Claude Mythos after Anthropic deemed it too dangerous to release, publishing an evaluation warning the model could hack into previously secure critical infrastructure and finding a sharp rise in cyberattack capability versus prior frontier systems; it published its inaugural Frontier AI Trends Report drawing on two years of evaluations, and further announced a refocus in June 2026 under Secretary of State Peter Kyle toward protecting citizens from AI misuse against institutions and democratic values.

Additional Intelligence Signals

Patent citation lineage, earnings-call mentions, and federal contract disclosures — automatically collected, not yet visible anywhere else on the site.

Sources

Every claim traced to a primary source — evidence, recent activity, and insider filing behavior, all in one place.

Evidence

2 sources · 2 government/SEC filings · 49% avg. source confidence

  • AI Security Institute (AISI)UK AI Security Institute evaluates Claude Mythos, warns of critical infrastructure hacking risk52% source confidence
  • ADS Group / RaconteurUK AI Security Institute announces refocus, publishes inaugural Frontier AI Trends Report46% source confidence

Correlated Activity

Both UK AI Security Institute (AISI) and US Center for AI Standards and Innovation (CAISI) which it depends on — are showing accelerating Activity at the same time (2 recorded changes for US Center for AI Standards and Innovation (CAISI) in the last 90 days).

Timing correlation only — not evidence of a causal link

Dependency Map

What UK AI Security Institute (AISI) depends on below, and who depends on UK AI Security Institute (AISI) above — click any node to make it the new center, 2 levels deep.

Connection type

Investment
Dependency
Supply Chain
Development
Enablement
Competition / Other
Succession

Click any node to make it the new center. Scroll to zoom, drag to pan.

The Story So Far (last 6 months)

Apr 2026: UK AI Security Institute evaluates Claude Mythos, warns of critical infrastructure hackin…Jun 2026: Announces refocus toward protecting institutions and democratic valuesJun 2026: UK AI Security Institute announces refocus, publishes inaugural Frontier AI Trends Report

Auto-generated from this entity's dated milestones, relationship updates, and sourced evidence — not AI-written, just sorted.

UK AI Security Institute (AISI)'s Timeline

A sourced, dated history of UK AI Security Institute (AISI)'s key moments — founding to present.

  1. Jan 2025 · Rebranded as AI Security Institute

    The Institute was rebranded from the AI Safety Institute to the AI Security Institute, emphasizing protection of citizens, institutions and democratic values from AI misuse.

  2. Jan 2026 · Evaluates Claude Mythos, finds sharp rise in cyberattack capability

    The UK AI Security Institute was given access by Anthropic to evaluate Claude Mythos, a model Anthropic deemed too dangerous to release, publishing an evaluation warning the model could hack into previously secure critical infrastructure and finding a sharp rise in the model's cyberattack capability compared with prior frontier systems.

  3. Jun 2026 · Announces refocus toward protecting institutions and democratic values

    Secretary of State for Science, Innovation and Technology Peter Kyle announced a refocus of the Institute's mission in late June 2026 toward protecting citizens from those who would use AI against institutions, democratic values and the UK's way of life, alongside publishing its inaugural Frontier AI Trends Report drawing on two years of evaluations.