Technology minister Kanishka Narayan warned UK firms that Anthropic’s new AI model, Claude Mythos Preview, can locate thousands of software vulnerabilities, raising fresh cyber‑risk concerns. The San Francisco start‑up disclosed the model’s ability to flag flaws in widely used applications, prompting alarm among regulators and corporate security teams worldwide.

Narayan said the government aims to turn the threat into a growth market, leveraging the £500mn Sovereign AI fund to bolster domestic defence capabilities. The unit will take a right‑of‑first‑refusal on start‑up investments, offering access to computing power and public datasets in exchange for commercial terms that give taxpayers a return. James Wise, chair, framed it as a novel financial instrument.

Anthropic has limited Mythos to a handful of large tech firms, but Narayan urged the National Cyber Security Centre to accelerate protection of critical infrastructure and raise private‑sector awareness. Wise confirmed that AI‑security will be a top priority for the Sovereign AI unit, which also promises fast‑track visas for talent. The initiative positions Britain as a hub for AI‑defence startups.

With Anthropic announcing a major UK office expansion, the move underscores Britain’s bid to retain AI expertise and monetize emerging security services.