Outsiders breached Anthropic’s Claude Mythos same day ‘potentially dangerous’ AI model was revealed: report
By Thomas Barrabi
Published on April 22, 2026.
A handful of users managed to gain unauthorized access to Anthropic's Claude Mythos, an AI model that the company claims is potentially dangerous and could cause a wave of cyberattacks if made available to the public. The breach occurred on April 8, the same day Anthropic and its CEO Dario Amodei revealed that Mythos was only available to about 40 handpicked corporate clients as part of "Project Glasswing". The unauthorized users, who belong to a private online forum dedicated to cracking unreleased AI models on Discord, have been using Mythos regularly, not for cybersecurity purposes. Despite attempts to restrict access, Anthropic has stated that there is no evidence that the unauthorized access has impacted any other systems.
Related Articles
Thailand Regulator Eyes Crypto Futures Expansion in Rule Proposal
Thailand's SEC is seeking public comment on proposed changes allowing digital asset businesses to apply directly for derivatives licenses, enhancing investor safety and compliance.
New Executive Director Named At Sag Harbor Cinema
The Sag Harbor Cinema Arts Center appointed Mark Lubell, a distinguished executive and cultural entrepreneur with a significant track record, as its executive director.
How Businesses Can Build Trust as Institutional Confidence Erodes
Edelman's Trust Barometer highlights declining trust among institutions, urging leaders to prioritize authentic leadership and mutual understanding amidst increasing class divide.