Anthropic Details Multi-Layered AI Code Review Process for Claude Code
Anthropic has shared how its engineering team manages software development using AI-driven reviews, automated verification, and risk-based human oversight.
🌐Work & SocietyPhoto: infoq.com
Key points
- Anthropic outlines multi-layered AI verification and risk-based human oversight for Claude Code.
- Merged pull requests doubled between June and September while reverted changes fell over 40 percent.
- Human approval requirements scale based on code risk, ranging from zero approvals for documentation to dual sign-offs for sensitive areas.
AI firm Anthropic has detailed the internal code review process utilized by its Claude Code team, highlighting a shift toward automated verification and risk-based human approval as AI agents take on a larger share of coding tasks. According to engineering leads Addy Osmani, Morgan Westlee Lunt, and Qing Wang, the approach is designed to cut down manual review time while keeping production incidents and code regressions in check.
Between June and September, merged pull requests within the Claude Code repository more than doubled, while the proportion of changes needing a revert dropped by over 40 percent. Rather than traditional line-by-line human inspection, verification is pushed earlier into the development pipeline. Before a pull request is even opened, Claude executes builds, tests, and linters, while a separate agent reviews the modifications in a fresh context.
Every pull request subsequently passes through automated checks where multiple AI reviewers independently investigate potential bugs. A dedicated security review can block merges if unresolved vulnerabilities persist. Meanwhile, human involvement is scaled according to risk: low-risk adjustments like documentation updates may require zero human approvals, medium-risk updates need one, and high-risk or sensitive changes require two or more oversight steps.
To prevent infinite feedback loops, automated review cycles are capped at three rounds before developers must rethink the implementation. Anthropic noted that internal pull requests passing after a single review and fix round increased from 65 percent to 78 percent under this methodology, whereas pull requests requiring five or more rounds fell from 8.9 percent to 1.5 percent. Human developers retain ultimate accountability for product direction, architectural choices, and high-risk modifications.
Source infoq.com — infoq.com/news/2026/11/claude-code-review/?utm_campaign=infoq_conte… ↗
Read next
⚖️Regulation & PolicyOpenAI Partners with Chilean Ministry to Advance Public Sector AI
OpenAI has signed a non-binding agreement with Chile's science ministry to integrate artificial intelligence into public administration and support local startups.
🧠AI AnalysisAI Models Vulnerable to Manipulation in Simulated Warfare Scenarios, Research Shows
New research reveals that leading AI models from firms like OpenAI and Anthropic can be manipulated into identifying civilian targets when prompted with game scenarios.
⚖️Regulation & PolicyAWS Highlights Enterprise Governance and AI Workspace Quick Ahead of re:Invent 2026
Amazon Web Services is spotlighting its workplace AI platform, AWS Quick, ahead of its upcoming re:Invent 2026 conference, emphasizing secure deployment across major ente…
🏢Big TechLoanPro Integrates Amazon Bedrock to Automate Loan Servicing Workflows
LoanPro has deployed a multi-agent AI assistant using Amazon Bedrock and Anthropic's Claude models to help lenders manage borrower accounts, resulting in a 15% reduction …
🏢Big TechGoogle Brings Wallet App to iOS Users Across the United States
Google has expanded the availability of its Wallet application to iOS devices in the U.S., allowing Apple users to store digital passes, tickets, and receipts.
⚖️Regulation & PolicyMicrosoft Reveals Cause of Global Outlook Outage
A latent software bug caused a nearly 24-hour global outage of Microsoft’s Exchange Online services, affecting millions of users worldwide.