Anthropic Adds Automated Security Reviews to Claude Code Amid AI-Generated Vulnerability Concerns

Aug 5, 2025

Anthropic has added automated security reviews to its Claude Code platform, addressing concerns that security practices can't keep pace with AI-generated code. The new features include a terminal command and GitHub integration to find vulnerabilities early. The launch comes as developer trust in AI code remains low and competition with OpenAI and Meta intensifies, positioning security as a key differentiator for Anthropic.

Claude Code security features

▪The new features include a "/security-review" terminal command for on-demand scans and a GitHub Action for automated reviews of pull requests.
▪The security scanner detects common vulnerabilities such as SQL injection, cross-site scripting (XSS), authentication flaws, and insecure data handling.
▪Anthropic has launched automated security review capabilities for its Claude Code platform to scan for vulnerabilities and suggest fixes.

AI-generated code trust issues

▪The same 2025 survey revealed that 46% of developers distrust the accuracy of AI-generated output, while only 3% reported a high level of trust.
▪The security tools address the problem of traditional manual security reviews being unable to keep pace with the exploding volume of AI-generated code.
▪A 2025 Stack Overflow survey found that 84% of developers are using or plan to use AI in their development workflows, up from 76% in 2024.

Automated security review integration

▪The GitHub Action integration automatically scans code changes and posts inline comments with recommended fixes directly in pull requests.
▪Anthropic tested the security tools on its own codebase, where they identified a remote code execution vulnerability and a Server-Side Request Forgery (SSRF) flaw.

Shift-left security practices

▪The tools are designed to accelerate "shift-left" security practices by embedding security earlier in the software development life cycle (SDLC).
▪The automated tools could democratize security practices for smaller development teams that may lack dedicated security personnel.

GenAI coding tool competition

▪The update was released one day after Anthropic launched Claude Opus 4.1, a more powerful AI model with improved coding capabilities.
▪The security feature launch comes amid industry competition, with OpenAI expected to announce GPT-5 and Meta recruiting AI talent with large bonuses.

Enterprise AI security risks

▪Analysts warn that a key risk of enterprise AI security tools is confusing fluency with accuracy, which can create a false sense of security.
▪Threat researchers from Cymulate previously discovered two high-severity vulnerabilities in Claude Code that allowed for arbitrary command execution.
▪According to Sanchit Vir Gogia of Greyhound Research, LLM-based tools like Claude Code can offer well-articulated but factually incorrect conclusions.