Security compliance you can prove, even when AI agents do the work.
I am a security compliance specialist: ISO 27001, SOC 2, ISO 42001 and the EU AI Act, from gap analysis to audit. I help security and compliance leaders keep their controls and evidence solid as AI agents join the work.
My working rule: agents will be wrong, so build so it doesn't matter. I run my own work with a fleet of AI agents under approval gates, isolation and evidence, and bring that method to governance, risk and compliance. The projects and guides here show it in practice.
Gate lab a small simulation
- 🤖Agent
- 🔍Review
- 🛡Scan
- 👤Owner
- 🏁Core ▣
- Agent proposes a change
- Review: reads the exact commit, looks fine
- Scan: no organisation data found
- Owner: approved the merge
- Core: merged, record kept
1merged0mistakes contained
Review is bound to the exact commit; the scan blocks organisation data; the owner approves the merge. Nothing here is real data.