Alberta uses Claude Code to review government systems
Claude Code helped Alberta identify vulnerabilities, generate fixes and add security review agents.
The Government of Alberta has used Anthropic’s Claude Code to review and secure provincial government systems, according to a case study published by the company. Anthropic said Alberta’s Ministry of Technology and Innovation used Claude Code with its Opus and Sonnet models to analyse code, identify vulnerabilities and support remediation.
According to Anthropic, the ministry scanned 466 million lines of code in about 20 hours, covering systems used across 27 provincial ministries. Around 50 AI agents worked in parallel to identify security vulnerabilities, infrastructure weaknesses and documentation gaps.
The ministry manages about 1,280 applications and 3,400 code repositories supporting services including social services, public safety and wildfire response. Anthropic said many had never undergone comprehensive security reviews, resulting in accumulated technical debt and incomplete documentation.
Alberta used a two-stage review process. A rules engine first identified known patterns, after which Claude Code analysed the results and cited the relevant files and lines for each finding. Anthropic said the approach uncovered issues that conventional automated scanning tools had missed.
Claude Code was also used to generate fixes, write tests where needed and assist with modernising legacy systems. Anthropic said ministry engineers reviewed and approved all proposed patches before deployment, maintaining human oversight throughout the remediation process.
Alberta also developed specialised Claude-based review agents for continuous security testing during software development. These include red-team agents that probe applications for vulnerabilities, blue-team agents that assess compliance with security standards, and additional agents that review code quality and public-facing content.
Why does it matter?
The case illustrates how governments are beginning to use AI coding agents to modernise and secure large portfolios of legacy software, an area that has traditionally required significant time and specialised expertise. If these tools prove reliable, they could help public administrations reduce technical debt, improve cybersecurity and accelerate software maintenance across critical public services.
At the same time, the deployment highlights the importance of governance in public-sector AI adoption. Alberta’s reported use of human review before implementing AI-generated changes reflects a growing emphasis on combining AI-assisted development with oversight, accountability and established security practices.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!