Breaking

UK Says It Will Work With Allies on AI Safety After Frontier Slowdown Warnings

September 15, 2026 at 06:00 EDT

Policy & Regulation
Foundation Models
Security

The UK government said it is prepared to work with international partners on AI safety after senior AI figures called for a slower pace of frontier-model development. First Secretary Louise Haigh said Britain would prioritize public safety and national security while still pursuing AI benefits for public services, business and science.

Sep. 15, 2026 · UK AI safety

Britain shifts toward AI safety talks

The government says it will work with partners after frontier AI leaders warned that model progress is outpacing safeguards.

Cyber capability is climbing

UK testing found apprentice-level cyber task completion rising from just over 10% in early 2024 to 50% in current evaluations.

Just over 10%Early 2024

50%Current evaluations

6-12

months in the stated risk horizon

steps in the pacing framework

reported code output per engineer

>80%

merged code authored by Claude

The proposed safety path

1. Embedded evaluators

Independent teams get ongoing access inside frontier labs.

→

2. Common standards

Democratic-country labs coordinate on verifiable safety bars.

→

3. Global coordination

Governments test what can be verified across borders.

Support

Outside access could make safety claims more verifiable after real-world cyber-evaluation failures.

Concern

Broad slowdown rules could become hard to enforce or favor the largest frontier labs.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.