US Restores UK Access to Test Latest AI Models After Month-Long Halt

The United States has restored the UK AI Safety Institute’s ability to test the latest artificial intelligence models from US companies, according to reports from Techmeme. This decision reverses a restriction imposed last month that halted the sharing of these advanced systems for safety evaluations.

Sources cited by the Telegraph indicate that US officials previously ordered OpenAI and Anthropic to stop sharing their models. This action effectively blocked UK regulators from reviewing the frontier models developed by these firms. The restriction also extended to models from Google DeepMind, limiting the scope of cross-border safety assessments.

Regulatory Shift on Model Sharing

The previous halt in data and model sharing created a significant gap in the UK’s ability to independently assess US-developed AI systems. By restoring access, the US government is allowing the UK AI Safety Institute to resume its safety testing protocols. This move signals a shift back toward cooperation on AI safety standards between the two nations.

The restoration of access applies specifically to the "latest US models," as noted in the reporting. This precision suggests that the focus remains on the most current and potentially powerful AI developments, rather than older or less capable versions. The specific models involved are those from the key US players in the frontier AI space.

Impact on Safety Testing

The UK AI Safety Institute relies on access to these models to conduct its safety evaluations. Without this access, the institute could not perform its mandated safety checks on systems developed by OpenAI, Anthropic, and Google DeepMind. The previous block meant that regulators were cut off from reviewing these specific frontier models during the hiatus.

This development is relevant for understanding the regulatory landscape of AI safety. The restoration of access allows for continued safety testing of frontier AI models. Businesses should monitor how these regulatory changes might affect the AI tools they use or develop.

Defender's checklist

The restoration of access allows for continued safety testing of frontier AI models. This is a positive step for maintaining robust safety standards in the AI sector. Businesses should monitor how these regulatory changes might affect the AI tools they use or develop.

What to do and how to stay safe: OpenAI

• Review your current AI vendor contracts to understand data sharing and safety compliance terms.
• Monitor regulatory updates from both US and UK authorities regarding AI safety standards.
• Ensure your internal AI usage policies align with the latest safety testing requirements.
• Stay informed about which specific models are subject to these safety testing protocols.

Step-by-step guide: Implementing AI Usage Policies at Work: A Step-by-Step Framework

General security guidance from the InfoSec Ledger newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Which companies were affected by the US block on model sharing?

The US blocked model sharing from OpenAI, Anthropic, and Google DeepMind, preventing UK regulators from reviewing their frontier models.

What has changed with the UK AI Safety Institute’s access?

The US has restored the UK AI Safety Institute’s access to the latest US models for safety testing, reversing the previous block.

When was the block on model sharing imposed?

The US ordered OpenAI and Anthropic to stop sharing models last month, as reported by the Telegraph via Techmeme.