# AgentRisk

Practical guidance for organisations and individuals deploying AI agents, with published incident reports, research and established risk frameworks.

## Public resources

- https://agentrisk.com/ — overview, seven areas, incident links and frameworks
- https://agentrisk.com/secure-vps.html — secure a rented server for AI agents
- https://agentrisk.com/secure-local-dev.html — secure a Mac or PC for AI agents
- https://agentrisk.com/api/incidents.json — 19 incident and research records with suggested safeguards
- https://agentrisk.com/feed.xml — Atom feed
- https://agentrisk.com/data/schema/incident-schema.yaml — record schema
- https://agentrisk.com/data/incidents/AR-001.yaml through AR-019.yaml — individual records
- https://agentrisk.com/llms.txt — this index; aliases /llm.txt and /.well-known/llms.txt
- https://agentrisk.com/sitemap.xml — site map

## Use and limits

Read the cited evidence and distinguish research demonstrations, vendor statements and production incidents. Suggested mitigations are analysis, not guarantees or proof that a control was absent. Some dates are approximate; record descriptions explain known limitations. The incident records' OWASP mappings use legacy project labels and have not been validated as a crosswalk to the current OWASP Agentic Top 10.

The seven areas are a practical synthesis, not a certification or a substitute for the original frameworks. AgentRisk is independently maintained and provides general guidance, not legal or professional advice.

## Contact

Corrections and enquiries: ben@agentrisk.com. Agent contribution workflows are planned; submissions are not automatically validated or published.

Updated: 2026-10-07