The corporate has invited roughly 150 extra organizations to Mission Glasswing.
When Anthropic introduced Claude Mythos at the beginning of April, the corporate did not launch the brand new frontier mannequin to the general public. As a substitute, it a made a preview accessible to pick organizations by means of an effort it referred to as Project Glasswing. “We fashioned Mission Glasswing due to capabilities we have noticed in a brand new frontier mannequin skilled by Anthropic that we consider might reshape cybersecurity,” the corporate wrote on the time. “AI fashions have reached a stage of coding functionality the place they will surpass all however probably the most expert people at discovering and exploiting software program vulnerabilities.”
Anthropic introduced Mission Glasswing alongside practically 50 companions, together with Apple, Amazon, Broadcom, CrowdStrike, Microsoft and NVIDIA. At this time, the corporate is inviting one other 150 organizations to the trouble. Anthropic says the invitees are based mostly in additional than 15 international locations and hail from industries that weren’t well-represented in Glasswing’s authentic cohort. Suppose: public utilities like energy, water and telecom firms, in addition to healthcare suppliers. Every group might want to move Anthropic’s safety necessities earlier than they’re granted entry to Claude Mythos.
“What every accomplice has in frequent is {that a} profitable assault on their codebase may very well be catastrophic,” the corporate notes. “For many companions, we estimate {that a} main assault might have an effect on greater than 100 million individuals, with essential ramifications for each world and nationwide safety.”
Anthropic says it is working “as shortly as we will to securely launch Mythos-level capabilities” to the general public, however provides it will not make the mannequin broadly accessible till it implements “extremely sturdy safeguards” to forestall potential misuse of its cyber capabilities. In Could, the announcement of Claude Mythos appeared to make the Trump administration rethink its stance on AI regulation. The president was anticipated to signal an govt order that might have established a course of by which the federal authorities would have had the means to share details about software program vulnerabilities recognized by AI methods with vital establishments earlier than these fashions have been made publicly accessible. Nevertheless, the signing ceremony was reportedly postponed on the final minute because of objections from Trump himself and the tech industry.
