Topic analysis

Anthropic investigating claim of unauthorised access to Mythos AI tool

Anthropic is investigating reports of unauthorized access to its Claude Mythos, a powerful unreleased cyber-security AI model, through a third-party vendor environment; no evidence of malicious use of the model has been found to date. UK cyber authorities, including the head of the National Cyber Security Centre (NCSC) and Security Minister Dan Jarvis, have highlighted both the risks of uncontrolled frontier AI access and its potential net positive impact on security, urging improved basic cyber practices and collaboration between AI firms and government.