Anthropic has officially launched Claude Security, now available in public beta for Claude Enterprise customers. This tool, developed by Anthropic, leverages the advanced capabilities of Claude Opus 4.7 to scan code for vulnerabilities and recommend patches. Anthropic highlights that the model goes beyond relying on known patterns; instead, it analyzes how code components interact across various files and modules, offering a deeper understanding of potential weaknesses. The scan results provide critical information, including severity, reproducibility, and confidence scores.
The public beta introduces several new features designed to enhance security operations, such as scheduled scans, CSV export functionality, and seamless integrations with popular collaboration platforms like Slack and Jira. Furthermore, key partners in the cybersecurity industry, including CrowdStrike, Palo Alto Networks, SentinelOne, and Wiz, are actively integrating Opus 4.7 into their own security products, aiming to bolster collective defense capabilities.
This launch comes in the context of recent developments where Anthropic's own “Mythos” AI model was deemed too dangerous to release due to its potent offensive capabilities, which an independent review by a British cybersecurity agency confirmed to be effective primarily against weak, unprotected networks. With OpenAI's recently released GPT-5.5 also demonstrating similar offensive prowess, Claude Security is positioned to equip defenders with the same cutting-edge AI advantage that attackers currently possess, helping to level the playing field in the evolving landscape of cyber warfare.