Some Unknown Group Is Reportedly Using Claude Mythos Without Permission

Anthropic, an artificial intelligence company, has confirmed that an unauthorized group has gained access to its highly sensitive AI model, Claude Mythos, which the company has described as too dangerous for public release. The breach reportedly occurred through a third-party vendor environment, with the group exploiting a combination of a data leak from AI startup Mercor and insider access from an employee at an Anthropic contractor. The group has allegedly been experimenting with Claude Mythos since early April, coinciding with the announcement of Anthropic’s Project Glasswing. The unauthorized access was revealed after the group shared a live demo and screenshots of Claude Mythos with Bloomberg, which subsequently verified the breach. According to an anonymous member of the group, their intent is exploratory rather than malicious, focusing on “playing around” with the model rather than causing harm. The group reportedly used internet sleuthing tools and bots on platforms like Discord and GitHub to locate the model’s online presence, leveraging leaked data and insider knowledge to bypass security measures. This incident raises significant concerns about the security of cutting-edge AI technologies, especially those deemed too powerful or dangerous for open use. Claude Mythos is considered one of the most advanced and potentially risky AI models, trusted only to Anthropic and its controlled environments. The breach highlights vulnerabilities in the supply chain and third-party vendor management, underscoring the challenges of safeguarding sensitive AI assets in a complex ecosystem. The implications of this unauthorized access extend beyond Anthropic, as it signals potential risks for the broader AI industry regarding data security and ethical control. While the group claims no malicious intent, the incident prompts urgent questions about how AI companies can better protect their models from exploitation and ensure responsible stewardship of powerful technologies. Anthropic’s ongoing investigation will be closely watched as the company works to assess the full scope and impact of the breach.