Anthropic Launches "Code Review" To Fix AI Code Security Issues

The rapid rise of AI‑generated code has left many organizations scrambling to maintain quality and security. Anthropic’s new Code Review tool, built into its Cloud Code platform, directly tackles this problem by automatically scanning every pull request that originates from AI assistants such as Claude or Codex. By flagging risky patterns before they reach production, the service promises to cut the volume of buggy releases that now plague enterprises where up to ninety percent of new code is machine‑written. This move reflects a broader industry shift toward AI‑assisted development paired with rigorous automated oversight.

Unlike many static analysis tools that concentrate on style or linting, Anthropic’s Code Review zeroes in on logical errors and potential security flaws. A multi‑agent architecture runs parallel analyses, each examining the code from a different perspective before a final aggregator removes duplicates and assigns severity colors—red for critical, yellow for warnings, and purple for legacy concerns. The system also offers a lightweight security scan and lets teams add custom checks aligned with internal policies. Pricing follows Anthropic’s token model, with typical reviews costing between $15 and $25, dramatically cheaper than hiring human auditors.

The launch arrives at a pivotal moment for Anthropic, whose enterprise subscriptions have quadrupled this year and whose Claude‑Code product already generates a $2.5 billion run‑rate. By embedding AI‑driven code review into Cloud Code, the company positions itself as a standard‑setter for large engineering organizations such as Uber, Salesforce, and Accenture. If the tool delivers on its promise of fewer bugs and faster releases, it could accelerate broader adoption of AI‑assisted development while raising the baseline for security hygiene across the software industry. Competitors are likely to follow, making automated logical‑error detection a new norm.