Anthropic experienced a multi-week security breach in which unauthorized users in a Discord group gained access to the Claude Mythos model. The incident was discovered and is now being publicly reported. The breach raises questions about access control and credential security in production AI systems.
Safety
Unauthorized Users in Discord Group Had Weekslong Access to Anthropic’s Supposedly-Super-Dangerous Claude Mythos Model
Weeks-long unauthorized access to Anthropic's Mythos model via a compromised Discord group exposes credential security gaps in production AI systems.
Thursday, April 23, 2026 12:00 PM UTC2 MIN READSOURCE: Daring FireballBY sys://pipeline
Tags
safety
/// RELATED
ProductsApr 21
Unauthorized group has gained access to Anthropic’s exclusive cyber tool Mythos, report claims
A vendor contractor's credentials were exploited to breach Anthropic's newly-announced Mythos cybersecurity tool on launch day, exposing supply-chain vulnerabilities in AI security products.
Safety5d ago
Copy Fail: 732 Bytes to Root on Every Major Linux Distributions
CVE-2026-31431 (Copy Fail) enables unprivileged users to achieve root access across all major Linux distributions via a 732-byte exploit script targeting kernel page cache corruption present since 2017.