Anthropic Broadens Access to Claude Mythos as IPO Filing Occurs

Anthropic Expands Claude Mythos Access Amid IPO Plans

Anthropic is increasing access to its Claude Mythos AI model through Project Glasswing, targeting technology firms and governmental organizations to help identify and fix software vulnerabilities before the model's public launch. This move comes as the company prepares for a potential initial public offering (IPO), having filed confidentially with the U.S. Securities and Exchange Commission.

On Tuesday, Anthropic revealed that it will include approximately 150 new organizations in its controlled access program, enhancing the rollout of this advanced AI technology. The firm anticipates making these Mythos-class models available to clients in the coming weeks, contingent on finalizing additional safety protocols.

“After several weeks of intensive collaboration with our Project Glasswing partners, we’re extending our partnership to around 150 new entities,” Anthropic stated. The expansion includes key sectors such as power, water, healthcare, and communications, where potential software breaches could have dire consequences for millions of users.

Previously introduced in March, Claude Mythos has already aided prior partners in discovering over 10,000 significant software vulnerabilities. However, its capabilities have raised concerns among researchers and governmental bodies regarding the potential for advanced AI systems to facilitate cyberattacks. For instance, during testing, Mythos autonomously conducted a simulated attack on a corporate network.

Despite the expansion, users on the Myriad prediction market express skepticism about the public release timeline, currently estimating only a 26.5% chance of Claude Mythos being available by the end of June, a noticeable drop from 58.5% just last week.