Anthropic opens Claude Security: AI that looks for flaws in code

Anthropic has released Claude Security as a public beta for Claude Enterprise customers. It is a tool that scans a company's code for security holes and suggests how they can be closed.

The tool was previously called Claude Code Security. It was first shown in February and was for a period only available to a limited number of users. According to the sources I have found, hundreds of organizations tried it during the closed test.

How it works

Many traditional security tools look for known patterns and fixed rules in the code. Claude Security works differently. According to coverage, it reads the code in context, just as a human security researcher would. It looks at how different parts of the program connect across files and follows how data moves around. The work uses Anthropic's Claude Opus 4.7 model, which also suggests fixes.

Companies can access the tool from the page in Claude.ai or at claude.ai/security. No separate programming or setup via API, which is technical access to the service, is required.

Features for large companies

A number of features that larger organizations typically need are built in:

- scheduled scans and scans of selected parts of the code - the option to dismiss a finding with a documented reason, so it can be traced during later audits - export of results as CSV and Markdown - webhooks that send messages to tools such as Slack and Jira

What are the results so far?

According to coverage, Anthropic found about five hundred vulnerabilities in production code during two months of private testing. Some of them had reportedly gone undetected for years, even after review by experts. The figures come from Anthropic's own information and have not been independently verified.

What does it mean for you?

As an ordinary user, you will not notice the tool directly. But if companies find and fix flaws in their software faster, it can reduce the risk of hacking and data leaks. At the same time, it is an example of AI increasingly being used to check code, not just to write it.

According to coverage, Anthropic plans to expand access to customers with Team and Max subscriptions.

What it means for you

If you or your workplace have Claude Enterprise, you can already try Claude Security via claude.ai/security and let it scan your code for security holes. It especially benefits developers and IT security staff, who are spared searching manually and get suggested fixes. If you work at a company that builds its own software, you can ask the IT department whether they have considered testing it. As an ordinary user, you can simply hope for fewer data leaks as flaws are found faster.