Anthropic kept its latest AI model out of UK government testing

In short: Anthropic did not provide its newest AI model to a UK government-backed testing group, a change that has worried some officials.

What happened

Anthropic, the company behind the Claude chatbot, withheld its latest AI model from the UK’s AI Security Institute, according to the Financial Times. The institute is a government-backed group that tests advanced AI systems for safety, similar to how a car might be crash-tested before being trusted on the road.

The report says this is the first time Anthropic has chosen not to share a new model with the UK testing effort. The decision has prompted concern inside parts of the British government, including fears that other big tech companies could also pull back from cooperation.

Some officials worry this could signal a more protectionist approach from tech groups. In simple terms, that means companies may prefer to keep their most important technology closer to home, and share less with foreign governments and outside testers.

Why it matters

These tests are meant to spot problems before powerful AI tools are widely used, such as systems giving harmful advice, leaking sensitive information, or being misused. If major AI companies stop providing access for independent checks, governments may have less visibility into how safe these systems are. For everyday people, that can affect how quickly and how confidently AI tools can be adopted in public services and business.

Source: Financial Times