Claude Security: The Public Beta That Revolutionizes Security

Claude Security: A Major Advancement in Vulnerability Detection

Anthropic recently announced the availability of its tool Claude Security in public beta for users of Claude Enterprise. This tool, which was initially launched in research preview under the name Claude Code Security in February, has been tested by several hundred organizations. Feedback from these users has been crucial in refining the product before its wider release.

Claude Security stands out for its ability to analyze codebases to identify potential vulnerabilities. Once these flaws are detected, the tool validates its findings before publishing them and offers specific fixes for each issue. Unlike traditional tools that focus on recognizing known patterns, Claude Security tracks data flows through files, reads the Git history, and understands the underlying business logic.

Each detected vulnerability is accompanied by a confidence level, a severity assessment, its likely impact, and steps to reproduce the problem. The proposed fixes can be reviewed in Claude Code on the web, allowing teams to work on them in context. However, human review and approval remain essential before any correction is applied.

New Features Introduced in the Public Beta

The public beta version of Claude Security brings several enhancements compared to the research preview phase. Among these new features is the ability to schedule scans at regular intervals, thereby automating checks. The tool also allows targeting specific directories for more precise scans.

Additionally, Claude Security offers the option to document the rejection of results, enabling future teams to rely on previous triage decisions. Scan results can be exported in CSV or Markdown formats, facilitating their integration into existing audit systems. Furthermore, thanks to webhooks, results can be sent to third-party tools such as Slack or Jira.

Availability and Access to Claude Security

Claude Security is now accessible in public beta for Claude Enterprise customers. Users can access it via the sidebar of Claude.ai or directly at claude.ai/security. Administrators have the option to enable the tool from the administration console. Anthropic also plans to extend access to Team and Max plans in the near future. Notably, no API integration or custom agent development is required to use the tool.

An Ambitious Cybersecurity Strategy for Anthropic

The launch of Claude Security is part of a broader strategy by Anthropic to strengthen its presence in the cybersecurity field. This initiative was recently initiated with the launch of Claude Mythos Preview, a model deemed too powerful for public release and reserved for about fifty partner organizations.

Claude Security is based on the Claude Opus 4.7 model, Anthropic's latest public model. This model incorporates cybersecurity safeguards from the Project Glasswing and features reduced cyber capabilities compared to Mythos Preview. Anthropic describes it as its most powerful public model in a blog post.

Finally, this launch allows Anthropic to promote its Opus model to its network of partners, including major players like CrowdStrike, Microsoft Security, and Palo Alto Networks.