Cybersecurity Concerns Over New AI Model
Access occurred on the same day Anthropic announced the controlled release of Mythos.
Mythos is part of Anthropic's Project Glasswing, allowing select organizations to use the Claude Mythos Preview model.
Hackers used a Discord channel and Mercor data breach info to guess Mythos's location.
Anthropic is investigating but found no evidence of system impact.
Group claimed intent was 'playing around,' but regulators worry about misuse.
4 days ago