Anthropic has opened Claude Security to public beta, making the AI-based vulnerability scanning and remediation product available to all Claude Enterprise customers.
The product can be accessed through the Claude platform or via integrations with a range of technology and service partners.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
The release brings features including scheduled and targeted scans, streamlined integration with audit systems, and support for tracking triaged findings. It also provides coverage for a broader set of organisational use cases.
With this development, customers can use Opus 4.7, Anthropic’s leading generally-available model for code vulnerability analysis, to review full repositories, specific directories, or code branches for software flaws.
No additional API integration or deployment of custom agents is needed to start using the product if the customer already has access to Claude.
Users initiate scans directly from the Claude.ai sidebar or via the designated security portal, selecting the target for analysis. At this point, Claude applies reasoning to understand component interactions as a security researcher would, tracing data flows and reading the underlying source.
During the scan, Claude does not rely solely on pattern recognition for known exploits but instead conducts a contextual analysis of code behaviour. The model examines how components interact across files and modules, identifies hard-to-detect issues, and provides detailed reports for every potential vulnerability.
Once the process is complete, users receive outputs that include an explanation of each finding, the confidence level in the result, severity assessment, likely impact, reproduction steps, and a recommended patch.
Anthropic said: “Detection quality is paramount. Teams have told us that high-confidence findings are what really accelerate security work.
“Claude Security’s multi-stage validation pipeline independently examines each finding before it reaches an analyst, which drives down false positives, and Claude attaches a confidence rating to every result. This means that the signal that reaches the team is worth acting on.”
Teams can address identified issues and open a Claude Code session on the web to work through proposed fixes in context. According to early enterprise users, the product allows for a rapid cycle from scan initiation to applied remediation.
Anthropic said: “Time from scan to fix is the metric that matters. Early users pointed to this consistently, with several teams going from scan to applied patch in a single sitting, instead of days of back-and-forth between security and engineering teams.”
Additional features enable teams to schedule scans for continuous monitoring, target specific directories within repositories, and dismiss findings while documenting reasons for audit history.
Findings can be exported in CSV or Markdown formats for integration with standard tracking and compliance systems, and scan results are compatible with notification and project management platforms such as Slack and Jira via webhooks.
Opus 4.7 powers Claude Security’s core functionality and is integrated into security solutions provided by partners such as CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, TrendAI, and Wiz. This extends AI vulnerability analysis within widely used enterprise environments.
Service partners including Accenture, BCG, Deloitte, Infosys and PwC are also collaborating with organisations to implement security workflows leveraging Claude’s capabilities.
The system includes safeguards, with mechanisms in Opus 4.7 that automatically detect and block requests related to prohibited or high-risk security applications. These measures are designed to provide defenders with advanced security features while restricting misuse.
Anthropic said that this release as part of its broader initiative to provide advanced cybersecurity tools to defenders, building on efforts such as Project Glasswing and the Cyber Verification Program.
The company has previously released Claude Mythos Preview, a system tested with a select group of partners for both vulnerability discovery and exploitation, as part of its research work under Project Glasswing.
Anthropic claimed that hundreds of organisations have participated in testing Claude Security since its initial research preview early this year, and feedback from those users has directly influenced the current version available in public beta. The AI start-up confirmed that future access will expand to more customers, including Claude Team and Max users.
Recently, reports emerged that Anthropic is set to secure an investment of up to $40bn from Alphabet. The deal involves an initial cash commitment of $10bn from Google, which values the Claude developer at $350bn.
