•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Reports that Anthropic’s latest and most powerful AI model, Mythos, was illegally breached have shaken parts of the tech industry, according to Euronews. The model has previously been described by Anthropic as “too dangerous to release publicly,” citing concerns about severe security vulnerabilities it could pose.
Bloomberg reports that an unauthorized group gained access to Mythos through the environment of a third-party service provider. The group is believed to be made up of members of a Discord channel that seeks information about unreleased AI models.
In a statement to Euronews Next, an Anthropic spokesperson said: “We are investigating reports of unauthorized access to the Claude Mythos Preview version via a partner's service environment.”
At present, there is no evidence that Anthropic’s core systems were affected, or that any activity extended beyond the third-party environment.
Despite the lack of evidence of wider impact, insider sources say the hackers tested sophisticated tactics to bypass technical barriers. After gaining control, the group reportedly used Mythos regularly for a period.
Before the incident, Anthropic had rolled out Project Glasswing, a limited-access testing program for leading technology and financial firms. The entities selected to access Mythos include Amazon, Apple, and JPMorgan Chase.
The push to test Mythos is also spreading across Wall Street. Banking groups including Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley are reportedly rushing to trial Mythos in their operational environments.
Mythos’s potential is described as extending beyond information processing to cybersecurity defense. In April, US Treasury Secretary Scott Bessent convened a meeting in Washington with senior bank executives, where the US government urged financial institutions to use Mythos as a core tool to probe and detect system vulnerabilities.
Because of its ability to identify security gaps with high precision, Mythos is characterized as a “double-edged sword.” If such capabilities are obtained by cybercriminals, it could be used to unlock vulnerabilities at scale.
The incident is being framed as a warning for AI companies about supply chain security. Even when core systems are robust, third-party connections can remain a critical vulnerability—potentially enabling dangerous technologies to leak through partner environments.
Source: Euronews.
Premium gym chains are entering a “golden era” that is ending or already in decline, as rising operating costs collide with shifting consumer preferences toward more flexible, community-based ways to exercise. Long-term memberships are shrinking, margins are pressured by higher rents and facility expenses, and competition from smaller, more personalized…
