Anthropic launches Claude Security beta

Anthropic has moved its security work from research preview into a customer product. The company has put Claude Security into public beta for Claude Enterprise users, packaging vulnerability scanning, validation and patch generation into a workflow tied to enterprise codebases. Anthropic says the service is designed to help security teams move from finding bugs to shipping fixes, rather than leaving findings in a queue. The launch comes as Anthropic continues to publish new data from Project Glasswing, its software security initiative built around Claude Mythos Preview. In an update published May 22, Anthropic said Claude Opus 4.7 had been used to patch more than 2,100 vulnerabilities in the three weeks since launch. In the same update, the company said AI is changing the bottleneck in software security from discovering bugs to verifying, disclosing and patching them. (anthropic.com) ### What exactly is Claude Security selling? Claude Security is Anthropic’s enterprise security product for scanning repositories, validating suspected vulnerabilities and proposing targeted fixes. Anthropic’s product page says the tool “reasons about your code like a security researcher” and is intended to close the loop between detection and remediation. ITPro, citing Anthropic’s launch materials, reported that Claude Security can scan full enterprise repositories and then open Claude Code to help fix the issues it identifies. (anthropic.com) That ties the product to Anthropic’s broader push to turn Claude into a set of workflow tools for developers and security teams, rather than a standalone chatbot. (anthropic.com) ### Where do the 10,000 and 2,100 numbers come from? Project Glasswing is the source of the larger vulnerability claims. Anthropic’s Glasswing page says Claude Mythos Preview has found thousands of high-severity vulnerabilities, including flaws in every major operating system and web browser. Anthropic’s May 22 research update says the company has now discussed “the early public evidence” of Mythos Preview’s performance after reports that the model surfaced more than 10,000 high- or critical-severity zero-day vulnerabilities across important software. (itpro.com) The 2,100 figure is narrower. Anthropic said that in the three weeks after launch, Claude Opus 4.7 was used to patch more than 2,100 vulnerabilities, and it contrasted that pace with slower open-source patching cycles that depend on volunteer maintainers. That figure refers to vulnerabilities patched with Claude-assisted tooling, not to the full set of flaws Mythos Preview has reportedly identified. ### Why is Anthropic keeping Mythos separate from Claude Security? (anthropic.com) Claude Mythos Preview is not being released as a general-purpose public model. Anthropic’s risk and research materials describe Mythos Preview as its most capable frontier model and say it has shown strong performance on cyber tasks, including finding and exploiting zero-day vulnerabilities under testing conditions. The U.K. (anthropic.com) AI Security Institute said in April that its evaluation of Claude Mythos Preview found continued improvement on cyber tests and significant gains on multi-step attack simulations. Anthropic’s current product strategy appears to separate that restricted model from the customer-facing security product, using Claude Security to expose defensive workflows without broadly releasing Mythos itself. That is an inference from Anthropic’s product and research pages. (www-cdn.anthropic.com) ### What else is Anthropic lining up around this launch? Anthropic launched Claude for Small Business on May 13, describing it as a package of connectors and ready-to-run workflows that place Claude inside tools used by small businesses. The company said the product is aimed at owners and lean teams that want automation inside existing software. TestingCatalog reported on May 23 that Anthropic is also preparing Mythos 1 for Claude Code and Claude Security, and that Claude Opus 4.8 is being readied for release in the coming weeks. (aisi.gov.uk) Anthropic has not announced those releases on its main news pages, but its help-center release notes confirm Opus 4.7 went generally available on April 16, giving a recent baseline for the company’s current rollout cadence. (anthropic.com) ### What should readers watch next? Anthropic’s next concrete milestones are already listed across its own product pages. Claude Security remains in public beta for enterprise users, Claude for Small Business is live, and Project Glasswing remains the company’s vehicle for publishing further Mythos-related security updates. TestingCatalog said Mythos 1 and Claude Opus 4.8 could arrive in the coming weeks, while Anthropic’s May 22 Glasswing update said the company would continue outlining what to expect next from Project Glasswing and from future Mythos-class releases. (testingcatalog.com) (anthropic.com)