The Extended Brief

White House reverses course and restores UK AI Safety Institute's access to frontier models

Brief by The AI News AI newsroom · Oct 13, 2026, 5:14 PM EDT edition

Original reporting by The Decoder — Manuel Uth · published Oct 12, 2026, 9:14 AM EDT

The UK's AI Security Institute can again test frontier models from Anthropic, Google DeepMind, and OpenAI early, ending a US-ordered cutoff that had blocked its independent evaluations.

Key points

• The White House restored the UK AI Security Institute's early access to models from Anthropic, Google DeepMind, and OpenAI. source ↗
• In September, the Trump administration had told the three companies to stop sharing models with the British agency. source ↗
• Diplomatic pressure from UK AI Minister Kanishka Narayan led to the reversal. source ↗

Practical applications

• Teams that reference UK AI Security Institute evaluations in procurement, compliance, or risk assessments can expect its frontier-model testing output to resume.
• Policy and government-affairs staff at AI labs should treat government access arrangements as reversible within months and plan engagement strategies accordingly.

Context

The UK AI Security Institute is a British government agency that evaluates advanced AI models, and it had early access to models from Anthropic, Google DeepMind, and OpenAI before the US ordered that sharing stopped in September. Early access arrangements let government evaluators test frontier models ahead of wider release. The reversal followed diplomatic pressure from the UK's AI Minister.

What to watch

• Watch for the institute publishing fresh evaluations of upcoming frontier models, which would confirm the restored access is operational.
• Watch whether the US imposes similar sharing restrictions on other countries' AI safety bodies.

Related briefs

• White House Says All AI Companies Must Report Security Incidents
• Democrats prepare to battle over AI regulation
• The U.S. Nuclear Plant Fleet Is Leaning into AI
• Subversion of clinical judgment by conversational artificial intelligence

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.