Alberta scans 466M lines of code with Claude in 20 hours

Anthropic published a case study (July 6, 2026) on how the Government of Alberta, Canada uses Claude Code to review and patch security vulnerabilities across the systems of 27 provincial ministries — roughly 1,280 applications and 3,400 repositories.

Key points

• About 50 Claude Code agents (Opus + Sonnet) ran autonomously and in parallel, scanning 466 million lines of code in ~20 hours — work the team estimates could otherwise take ~6.5 years.
• A two-stage routine: a rules engine flags known patterns, then Claude reviews and cites the exact file and line so developers can verify; it surfaced issues that traditional scanners missed.
• Where a vulnerability was found, Claude Code could generate a fix, write tests, and build; overly outdated code was rebuilt in a modern language (a subsidy portal hand-coded in Java ~25 years ago was rebuilt in 4–5 days).
• Alberta also built continuous "red team"/"blue team" review agents, checking each application against ~95 security controls; every patch was reviewed and approved by engineers before shipping.

Alberta has released technical white papers so other governments can learn from the approach.