Does Anthropic’s Claude Mythos break the cyber insurance underwriting model?

Anthropic has announced that it will not release its latest AI model, Claude Mythos, citing global cyber security concerns. The company also described Project Glasswing, an effort to use a preview version of the model to secure critical software.

The article highlights the potential impact on the cyber insurance industry. Insurers rely on understanding the capabilities and risks of AI models to underwrite policies. With Claude Mythos withheld, underwriters face uncertainty about the model's full capabilities and potential for misuse. A source quoted in the article notes that 'the uncertainty is uninsurable,' suggesting that the lack of public availability complicates risk assessment.

The implications extend beyond a single model. The article suggests that similar actions by other AI labs could reshape the cyber insurance market, forcing insurers to adapt to a landscape where powerful models are kept private or selectively released. This may require new methodologies for evaluating cyber risk in the absence of full public access to frontier AI systems.