Anthropic is reportedly on the verge of releasing Claude Mythos to the public, with reports suggesting the landmark AI model could go live as early as June 10. The release marks a significant milestone not only for the company, but for the wider business ecosystem, as the new model brings increased strengths and potential risk thanks to its reputed capabilities in cybersecurity.
Reported by a media outlet run by journalist Alex Heath, the public version will arrive under a new name to reflect the substantial guardrails and more limited cybersecurity capabilities than those available to existing partners. To understand the stakes: Anthropic estimates that a major attack on any one of its Glasswing partners’ codebases could affect more than 100 million people, a figure that underscores just how consequential the model’s capabilities and its release truly are.
The public rollout follows months of controlled access, with Anthropic having granted a select group of partners access to Claude Mythos Preview in early April. That preview period served as a proving ground for the model’s ability to identify critical vulnerabilities in some of the world’s most sensitive software, and the results have been striking.
What makes this release particularly notable is not just the model’s technical capability, but the deliberate, collaborative approach Anthropic has taken to bringing it to market. Rather than a standard product launch, the rollout has been shaped by ongoing dialogue with governments, security researchers, open-source maintainers, and critical infrastructure operators, making this one of the most structured model deployments the AI industry has yet seen.
What Project Glasswing Has Already Uncovered
The data emerging from Project Glasswing‘s preview phase tells a compelling story. Since early April, the initial cohort of roughly 50 partners, which included the U.S. government, has used Claude Mythos Preview to scan their codebases for vulnerabilities. Their collective findings have been substantial: more than 10,000 high- or critical-severity security flaws identified in just a matter of weeks.
Now, Anthropic is scaling significantly. The company last week announced an expansion of Project Glasswing to approximately 150 new organizations across more than 15 countries, with most providing critical infrastructure to many more. The expansion brings in sectors that were underrepresented in the original cohort, including power, water, healthcare, telecommunications, and hardware.
Many of the new partners are vendors, companies, and nonprofits whose codebases underpin the operations of countless other organizations globally, including governments. The common thread across all of them is exposure: a successful attack on any one of these partners could have cascading consequences at a scale that few other cyber incidents have ever reached.
To support the broadened program, Anthropic has also released Claude Security, a product built on its latest public frontier models, including Claude Opus 4.8, designed to scan codebases and suggest patches. Separately, it has begun providing trusted security teams with the tools developed to help Glasswing partners identify vulnerabilities more efficiently.
A Measured Launch in a High-Stakes Landscape
The arrival of Claude Mythos Preview earlier this year did not happen quietly. It ignited wide-ranging conversations across the software industry and within governments about how AI is fundamentally reordering the cybersecurity landscape. Those conversations have directly shaped how Project Glasswing has evolved and expanded.
Central to Anthropic’s thinking is a warning the company has been sounding for some time: within six to twelve months, it expects many other AI companies to have Mythos-class models. The concern is that those models could be released without the safeguards needed to prevent misuse, and in that scenario, the frequency and unpredictability of cyberattacks could increase dramatically.
That warning resonates with security professionals watching the space closely. Morgan Adamski, Principal at PwC, put it plainly when discussing the implications of these highly capable cybersecurity models like Claude Mythos:
“Everyone should really be preparing for a breach in the next two years.”
That’s because, according to tests by UK government organization AI Security Institute, Mythos was the first model to fully complete the TLO challenge end to end, though only in a minority of attempts. On average, it completed 22 out of 32 steps per run, compared with a lower baseline from earlier models such as Claude 4.6, which averaged around 16 steps.
The measured approach Anthropic has taken — controlled access, phased expansion, sustained industry collaboration — appears designed precisely to get ahead of that risk. The company has framed Project Glasswing not merely as a product initiative, but as an effort to establish new operating norms for an era in which powerful AI cyber capabilities are increasingly widely available.
What Comes Next for Mythos and the Broader Industry
The public release of Claude Mythos marks the burgeoning of the next AI age in cyber, one which may be harder to challenge than previous iterations. The company has been candid that robust safeguards capable of preventing misuse of the model’s cyber capabilities have yet to be fully developed, by Anthropic or, to its knowledge, any other player in the industry.
In the near term, Anthropic plans to continue expanding Project Glasswing, prioritizing essential infrastructure providers, maintainers of critical open-source software, and safety testers. The company also intends to scale its Cyber Verification Program, which would extend Mythos-class capabilities to a wider set of organizations for specific cyberdefense tasks.
Longer term, Anthropic has been direct about what lies ahead: frontier model releases will only grow in consequence. As AI capabilities improve across all domains, the challenge of deploying powerful, dual-use models responsibly will not get easier. But the company believes Project Glasswing has given it a framework for navigating these inflection points.
The imminent public launch of Claude Mythos is, ultimately, a test of whether that framework can scale, and whether the industry is ready to meet it.
