Connect once, scanned with every change
Anthropic said on Claude's official X account on Oct. 21 (Oct. 22 Japan time) that it had added a feature to "Claude Security," its code vulnerability scanning service for businesses, that continuously scans connected repositories. When repositories on services such as GitHub or GitLab are connected, it first examines them in detail and then automatically rescans according to the risk and complexity of changes. Multiple repositories can be grouped into one project, allowing it to search for flaws that span services.
Claude Security began trial availability in April, and thousands of organizations have used it so far, finding more than 1 million vulnerabilities, the company said. The new feature is also offered on a trial basis to customers of "Claude Enterprise," the business-oriented plan.
Fix verification also automated
Flaws it finds are checked against code changes, and when a fix is made, they are automatically recorded as "fixed." This applies whether the fix is made with Claude Code or by other means. If the same problem appears again, the record is automatically reopened.
For each project, users can write down how the system is deployed and the company's internal security policies. Claude summarizes the attack paths that can be expected from the code, and users can revise them. Scans use this to set priorities and skip risks that have already been addressed. For each flaw, it explains the severity and how an attacker could reach that part of the code.
Top model also used, $25,000 a month free for 2 months
Scans use all of Anthropic's models, including "Claude Mythos," its top model that is not publicly available. The review normally required to use models with high cyber capabilities is not needed. Enterprise customers get $25,000 worth of usage free each month for two months. This covers use through Dec. 31.
The company also provided an API for reading and writing results and a mechanism for notifying external systems of new flaws. It can also be called from Claude.ai and Claude Code, letting engineers check and fix flaws within a single workflow.
In early October, Anthropic also launched "OSS Scanner," which examines important open-source software for free. On the same day, OpenAI disclosed that it had found "PHP2Shell," a serious flaw in PHP, using its model.
