Anthropic Mythos AI dey under scrutiny after vendor-linked access breach

Anthropic talk say dem dey investigate wan alleged unauthorised access to dia controlled-release cybersecurity model Mythos AI (Claude Mythos Preview). Reports say di attacker enter di preview environment through third-party vendor setup, no be by directly compromising Anthropic’s core systems. Di exposure reportedly happen around di time Anthropic roll out Mythos AI under Project Glasswing. Anthropic talk say dia internal investigation no show evidence say dia main systems were impacted, and di access look like e limited to di preview wey dem access through vendor channels. Later reports add more context: Mythos AI dey positioned as offensive/defensive tool wey fit autonomously find vulnerabilities and run complex exploits. External evaluators like METR and UK AI Security Institute reportedly find say im performance pass or reach existing cyber benchmarks. To manage dual-use and deployment risk, Anthropic limit access to about 40–50 vetted organisations under Project Glasswing. Even with controls, di program reportedly suffer breach for di April 21–22 window, wey shift focus to supply-chain and third-party access controls. Anthropic also disclose security funding commitments: up to $100 million in usage credits and $4 million in direct open-source security donations. Crypto-trader takeaway: dis incident no really about one specific token but about operational security risk. For crypto infrastructure, “security” dey expand beyond smart-contract audits to key management, oracle systems, and vendor/supply-chain monitoring—because failures for Mythos AI-type deployments fit quickly erode enterprise trust and readiness to adopt AI tooling.