Thanks for your interest in contributing. This list aims to be the most comprehensive public reference for AI agent security incidents.
Every incident submission must include:
- Date - Exact date or month/year when the incident occurred or was disclosed
- Named entity - Specific company, product, or project affected (no anonymous or hypothetical incidents)
- Impact - Concrete, measurable impact (users affected, dollars lost, data exposed, systems compromised)
- Root cause - Technical explanation of what went wrong
- CVE - If one exists, include the CVE identifier and CVSS score
- Source link - At least one verifiable source URL from a reputable publication or official advisory
- Documented AI agent security incidents with real-world impact
- AI agent framework and tool vulnerabilities with published advisories or CVEs
- AI chatbot integration breaches affecting real organizations
- Supply chain attacks targeting AI development tools and libraries
- Research disclosures demonstrating exploitable vulnerabilities in production AI systems
- AI agent autonomous behavior incidents (unintended actions, sandbox escapes)
- Product pitches - Do not submit entries that promote any security product, tool, framework, or vendor solution
- Unverified claims - Every claim must have a verifiable source
- Opinions or commentary - This is a factual timeline, not an editorial
- Hypothetical attacks - Only incidents that actually occurred or vulnerabilities confirmed to exist
- Links to vendor marketing pages - Sources must be news articles, official advisories, CVE databases, or peer-reviewed research
- Duplicate entries - Check existing entries before submitting
- Use standard Markdown table format matching the existing tables
- Sort entries by date descending (newest first) within each year section
- Do not use em dashes (the long dash character) - use regular hyphens or rewrite
- Keep descriptions concise and factual
- Include CVE numbers where they exist
- Use short, descriptive link text for source URLs
- Fork the repository
- Add your entry to the appropriate year section in README.md
- Verify all source URLs are accessible
- Submit a pull request with a clear title describing the incident being added
- Every source URL must resolve to a real page
- Financial figures must match reported amounts from cited sources
- CVE numbers must match entries in the NVD or relevant advisory databases
- Technical details must be accurate and verifiable from the cited sources