Anthropic says Alberta used Claude to scan 466 million lines of government code

Anthropic says the Government of Alberta has been using Claude Code to review, repair and modernize software across provincial systems, turning a large public-sector security audit into a case study for AI-assisted cyber work.

In a July 6 case study, the company said a team inside Alberta's Ministry of Technology and Innovation used Claude Code with Opus and Sonnet models to assess 466 million lines of government code in about 20 hours. The review covered roughly 3,400 code repositories and 1,280 applications maintained for 27 provincial ministries, including systems tied to social services, public safety and wildfire response.

What Alberta says it did

According to Anthropic, Alberta's team ran about 50 agents in parallel. The workflow first used a rules engine to flag known issue patterns, then had Claude review those flags and cite specific files and lines so developers could verify the findings. The company says the process found vulnerabilities, infrastructure weaknesses and documentation gaps, including issues missed by traditional automated scanners.

The effort did not stop at reporting. Anthropic says Claude Code could often generate fixes, run builds and write missing tests before patches were reviewed by ministry engineers. In some older systems, the case study says Claude helped rebuild applications in more modern languages when patching the original code was less practical.

• Alberta's systems include sensitive records such as tax, procurement and social services data.
• The province built specialized review agents for red-team probing, blue-team defense checks, code quality and public-facing writing.
• Anthropic says each application is checked against roughly 95 security controls during the review process.

Why it matters

Governments often carry decades of technical debt in systems that are difficult to document, patch and replace. Alberta's reported use of Claude is notable because it applies coding agents to an operational security problem rather than a narrow demo: scanning legacy repositories, producing evidence for human review and folding checks into the development process.

The case study still represents Anthropic's account of the project, so the reported time savings and technical results should be read as company-supported claims. Even so, the numbers make the deployment a useful marker for how AI vendors are positioning code agents in cybersecurity: not only as developer assistants, but as tools for continuous review and modernization inside large public institutions.

Alberta has also published technical white papers for other governments, and Anthropic says the province plans to expand the approach across more ministries this fall.

Sources

• www.anthropic.com

Cover photo by Christina Morillo on Pexels, used under the Pexels License.