Anthropic’s Glasswing and $800M raise

Anthropic did two very different things in the same week: it kept a powerful new security model behind a gate, and it moved to put real money behind selling Claude into private-equity portfolios. Those are not separate stories. They are the same company deciding it wants to be closer to the work, not just the model. (anthropic.com) (cnbc.com) (thenextweb.com) The security side is called Project Glasswing, and Anthropic says it brings together Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks to secure critical software. Anthropic says the effort is built around an unreleased model called Claude Mythos Preview. (anthropic.com) Anthropic says Claude Mythos Preview has already found thousands of high-severity vulnerabilities, including flaws in every major operating system and web browser. A software vulnerability is a hidden crack in code that can let an attacker break in, steal data, or shut systems down. (anthropic.com) That is why Anthropic is not offering Mythos like a normal chatbot. In its April 8 announcement, the company said the model can surpass all but the most skilled humans at finding and exploiting software vulnerabilities, so access is being limited to defensive security work. (anthropic.com) (forbes.com) Project Glasswing is not just a press release with logos. Anthropic says it is committing up to $100 million in usage credits for Mythos Preview and another $4 million in direct donations to open-source security groups, while more than 40 additional organizations that build or maintain critical software infrastructure are getting access to scan both their own code and shared open-source code. (anthropic.com) Then there is the second move: Anthropic is in talks to invest about $200 million into a new private-equity-backed venture that could raise roughly $1 billion in total. Reports say firms in the discussions include Blackstone, General Atlantic, and Hellman & Friedman, and the venture would help portfolio companies deploy Claude. (msn.com) (investinglive.com) (techfundingnews.com) Private equity firms own or back hundreds of companies at once, so they can act like a giant distribution network. If Anthropic gets Claude embedded across those portfolios, it is not selling one software license at a time; it is trying to become the default operating layer across whole clusters of businesses. (cnbc.com) (thenextweb.com) The reported model looks less like a pure software subscription and more like a consulting arm. CNBC, citing The Information, said the venture would use a Palantir-style approach, which means pairing the model with hands-on implementation teams that actually rebuild workflows inside companies. (cnbc.com) Put the two announcements together and the pattern is clear. Anthropic is keeping its most sensitive cyber capability tightly controlled while building a commercial machine that can push Claude deeper into banks, manufacturers, healthcare companies, infrastructure operators, and other enterprise systems where private equity already has ownership and leverage. (anthropic.com) (cnbc.com) That is a shift from being a company that mainly sells access to a model through an application programming interface, which is a paid doorway other software uses to talk to Claude. It is turning into a company that also sells outcomes: fewer security holes on one side, and faster cost-cutting and workflow replacement on the other. (anthropic.com) (thenextweb.com)