Recent reports suggest an unauthorized group may have gained access to Anthropic’s new AI model, Claude Mythos Preview. Due to the model’s elevated capability to identify digital vulnerabilities, access to Mythos is limited to a select group of partners in an initiative known as Project Glasswing, and so the potential breach reported by Bloomberg News concerns some security experts. 

“Unauthorized users were able to access Anthropic’s Mythos model, reportedly by just changing a model name,” says Shane Fry, Chief Technology Officer at RunSafe Security. “Even if their intent is just to explore, it shows how easily these systems can be exposed.” 

According to the report, users in a private online forum accessed Mythos the same day the company announced a limited release of the model. 

Anthropic confirmed to CBS News that an investigation into them matter is underway. The investigation is looking into the possibility that the access originated from one of its third-party vendor environments, which the company collaborates with to develop its models. Currently, no breaches outside of the vendor environment have been identified. A compromise of Anthropic systems has also not been detected. 

“The reality is these AI capabilities are already out there, ‘hacked’ or not, and they’re going to accelerate how quickly vulnerabilities are found and exploited,” Fry concludes. “Software teams will need to look at how to harden their code so those vulnerabilities can’t be used in the first place.”