Anthropic's Mythos-class model allegedly leaked on Hugging Face

Anthropic's unreleased Mythos-class model has allegedly leaked on Hugging Face, according to a post on Hacker News by user handfuloflight, though the company has not confirmed the breach. The post, which appeared roughly 53 minutes before this report, has already had its linked URL deleted over malware concerns, complicating efforts to verify the weights' authenticity. A Mythos-class model would represent a significant escalation in Anthropic's frontier lineup, making any unauthorized release a potential competitive and security risk. The incident follows a pattern of high-profile model leaks that have roiled AI labs and raised questions about internal access controls. Watch for whether Anthropic issues a takedown request, confirms the leak, or discloses how the weights were exposed.