Anthropic has announced the public beta launch of Claude Security, a cybersecurity tool designed for its Claude Enterprise customers. The newly introduced tool, formerly known as Claude Code Security, utilizes the latest version of Anthropic’s AI model, Claude Opus 4.7, to scan software codebases for vulnerabilities and provide remediation solutions.
According to the company, Claude Security features both scheduled and targeted scans, streamlined integration with audit systems, and enhanced tracking capabilities for identified vulnerabilities. Notably, organizations using Claude can start utilizing the scanning tool immediately, without the need for API integration or custom agent builds.
Access to Claude Security for Claude Team and Max customers is expected to roll out soon. Users can access the tool via the Claude.ai sidebar or directly at claude.ai/security.
One of the key differentiators of Claude Security is its method of vulnerability detection. Rather than relying solely on known patterns to identify weaknesses, Claude analyzes code similarly to a human security researcher. It evaluates how different components interact across files and modules, traces data flows, and interprets the source code. The AI then delivers comprehensive explanations of each identified finding, detailing the likelihood that a vulnerability is genuine, its severity, potential impacts, and how it can be reproduced. Furthermore, it generates instructions for targeted patches.
During a limited research preview, Claude Security was tested by hundreds of organizations of varying sizes, leading to enhancements based on two months of feedback. Notable improvements include a multi-stage validation pipeline designed to independently assess each finding before it is presented to analysts, thereby reducing the occurrence of false positives. Claude assigns a confidence rating to each result to assist in triaging vulnerabilities.
Additional features allow users to schedule scans for regular vulnerability assessments, target scans at specific directories within a repository, and document reasons for dismissing findings to help future reviewers trust prior decisions. Users can also export findings in formats such as CSV or Markdown for integration with existing audit systems and send scan results to tools like Slack and Jira through webhooks.
The capabilities of Claude Opus 4.7 are also being integrated into various cybersecurity software applications already in use by many enterprises, including offerings from CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, TrendAI, and Wiz.
This launch comes shortly after Anthropic introduced Project Glasswing and the frontier AI model Claude Mythos Preview, which is currently available only to select partners. In testing, the Mythos model reportedly uncovered thousands of previously unidentified zero-day vulnerabilities.
In a parallel development within the industry, OpenAI has recently unveiled GPT‑5.4‑Cyber and expanded its Trusted Access for Cyber program, which aims to facilitate more permissive and streamlined deployment of AI models for cybersecurity defense applications.
As the landscape of cybersecurity evolves rapidly, tools like Claude Security represent significant advancements in the use of artificial intelligence for identifying and mitigating software vulnerabilities. The integration of AI into cybersecurity practices not only enhances the efficiency of vulnerability management but also signals a broader trend towards automated, intelligent solutions in the fight against cyber threats.
See also
80% of Cybersecurity Investors Plan Increased AI Spending by 2026, Focus on ROI
Anthropic Launches Beta of Claude Security AI Tools to Combat Cyber Threats
Anthropic’s Mythos Reveals Thousands of Vulnerabilities, Banks Prepare for AI Cyberattacks
Anthropic’s Claims of AI-Driven Cyberattacks Raise Industry Skepticism
Anthropic Reports AI-Driven Cyberattack Linked to Chinese Espionage
