"A small group of unauthorized users gained access to Claude Mythos Preview, Anthropic's closely restricted cybersecurity AI model, on the same day the company publicly announced the model's existence."
"Anthropic confirmed it is investigating the claims: 'We're investigating a report claiming unauthorized access to Claude Mythos Preview through one of our third-party vendor environments.'"
"The significance of the breach is inseparable from the nature of the model. Anthropic announced Mythos Preview and the accompanying Project Glasswing initiative on 7 April 2026."
"The company withheld the model from general release specifically because of its offensive cyber capabilities: in testing, Mythos autonomously discovered thousands of previously unknown zero-day vulnerabilities."
A small group accessed Claude Mythos Preview by guessing its URL on the day Anthropic announced the model. The breach underscores risks in restricting AI access through vendor environments. Anthropic is investigating the unauthorized access but found no evidence of impact on core systems. The group, communicating via a private Discord channel, has used Mythos and provided proof of access. An employee at a third-party contractor may have facilitated this breach, which is significant due to Mythos's offensive cyber capabilities.
Read at TNW | Anthropic
Unable to calculate read time
Collection
[
|
...
]