Anthropic delays Claude Mythos

Anthropic has postponed a broad public release of Claude Mythos and is keeping the model inside a limited cybersecurity program instead. (anthropic.com) In its April 7 system card, Anthropic called Claude Mythos Preview its “most capable frontier model to date” and said the model’s jump in capability led it to decide not to make it generally available. The company said it will use Mythos with a restricted group of partners while it studies safeguards. (anthropic.com) Anthropic launched that partner effort on April 7 as Project Glasswing, giving early access to Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Microsoft, Nvidia, Palo Alto Networks, the Linux Foundation, and more than 40 other organizations tied to critical software infrastructure. Anthropic said it is committing up to $100 million in usage credits and $4 million in donations to open-source security groups. (anthropic.com) The company’s explanation is straightforward: it says Mythos can find and help fix software flaws at a level that changes how quickly defenders need to act. Anthropic’s preview page said the model has shown it can reverse-engineer exploits in closed-source software and turn known-but-unpatched vulnerabilities into working exploits. (anthropic.com) That release choice lands in the middle of Anthropic’s push toward public markets. Bloomberg reported on March 26 that Anthropic was considering an initial public offering as soon as October 2026 and had held early talks with Goldman Sachs, JPMorgan Chase, and Morgan Stanley about senior underwriting roles. (finance.yahoo.com) Bloomberg’s report, republished by Yahoo Finance, said a listing could raise more than $60 billion and that Anthropic had been valued at $380 billion in a February funding round co-led by MGX. Reuters reported in December 2025 that Anthropic had hired Wilson Sonsini to prepare for a possible offering, while an Anthropic spokesperson said the company had not decided when or whether it would go public. (finance.yahoo.com, (finance.yahoo.com) Anthropic has spent the past year building a public case that safety process is part of the product, not a side document. Its Transparency Hub says the company publishes model reports, system cards, and deployment safeguards to explain what Claude models can do and how Anthropic is handling risk. (anthropic.com) That framing now has a concrete test: Mythos is not being sold as a normal chatbot upgrade, but as a restricted tool for defenders working on critical systems. Anthropic’s own system card says the findings from Mythos will be used to shape future Claude releases and their safeguards. (anthropic.com) For now, the model that could have been Anthropic’s next mass-market launch is staying behind a narrower gate. The company is still moving Mythos into real-world use, just with banks, cloud providers, software firms, and infrastructure maintainers instead of the public internet. (anthropic.com, (anthropic.com)