claude.com
|
ksl
|
|
Anthropic launched Claude Security in public beta for Enterprise customers, a vulnerability scanner built on Opus 4.7 that reads source code and traces data flows across files rather than relying on pattern matching. It generates patches, attaches confidence ratings, and includes a multi-stage validation pipeline to reduce false positives. Scheduled scans, directory targeting and webhook integrations with Slack and Jira are already built in. The partner list is telling – CrowdStrike, Wiz, Palo Alto Networks, and Microsoft Security are embedding Opus 4.7 into their own products. Between this, GPT-5.4-Cyber, and the Vercel breach weeks ago demonstrating how AI tool integrations create new attack surfaces, security is becoming the most contested vertical in the AI platform war.
