Anthropic investigating claims of rogue access to its high-risk Claude Mythos AI: Report
Anthropic is investigating claims that unauthorized users accessed its unreleased Claude Mythos AI model through a third-party vendor environment. The model, capable of identifying and patching software vulnerabilities, has raised concerns about AI safety and potential cyber threats.
Anthropic, a US-based AI startup, is investigating unauthorized access to its unreleased Claude Mythos AI model. The model was accessed through a third-party vendor environment, according to a Bloomberg report. Mythos is a large language model capable of processing software code and identifying vulnerabilities. Anthropic had limited access to Mythos to a closed group of collaborators, but the breach has raised concerns about AI safety. The UK's AI Security Institute had cautioned that Mythos posed significant cyber threats, citing its ability to orchestrate complex attacks. Anthropic has pledged to strengthen defensive cybersecurity capabilities with Mythos as part of its Project Glasswing.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.