Anthropic embeds AI-powered security scanning into its coding assistant, uncovering 500+ previously unknown flaws in live open-source projects and auto-suggesting human-approved fixes.
Anthropic has introduced Claude Code Security, a built-in AI feature inside Claude Code that scans entire repositories and generates patches for software vulnerabilities, positioning the tool as an automated security reviewer for open-source ecosystems.
The system analyses code through multi-stage verification, detects complex logic flaws that rule-based scanners often miss, and produces fix suggestions with severity grading and confidence scores. Findings appear in a unified dashboard, while every change requires developer approval.
Explaining the difference, the company said: “Existing analysis tools help, but only to a point, as they usually look for known patterns […] Rather than scanning for known patterns, Claude Code Security reads and reasons about your code the way a human security researcher would: understanding how components interact, tracing how data moves through your application, and catching complex vulnerabilities that rule-based tools miss,” — Anthropic.
Anthropic stressed governance controls: “Nothing is applied without human approval: Claude Code Security identifies problems and suggests solutions, but developers always make the call.”
The open-source impact is already tangible. Using Claude Opus 4.1, the company uncovered more than 500 previously undetected vulnerabilities in production open-source codebases and is conducting responsible disclosure with maintainers.
“We’re working through triage and responsible disclosure with maintainers now, and we plan to expand our security work with the open-source community,” Anthropic said.
The tool has been stress-tested in Capture-the-Flag events and evaluated with Pacific Northwest National Laboratory for critical infrastructure defence. The move follows growing risks flagged by Tenzai, which warned that AI-built apps from platforms such as OpenAI, Cursor, Replit, and Devin can leak data or mishandle funds.
Internally, adoption is deep. Mike Krieger, Chief Product Officer, said:
“Claude is being written by Claude. Claude products and Claude code are being entirely written by Claude.”
