White paper WP-015
Cryptographic Audit Trails for Autonomous AI Agents
Mutable logs cannot keep up with agents that act faster than your reviewers.
At a glance
- Paper
- WP-015
- Topic
- Cryptographic Audit Trails for Autonomous AI Agents
- Format
- PDF + web summary
- Signatures
- ML-DSA-65 post-quantum (NIST FIPS 204)
- Sandbox
- Reproducible at affix-io.com/sandbox
- Company
- AffixIO, Wales, UK
Cryptographic Audit Trails for Autonomous AI Agents is an AffixIO technical paper. Mutable logs cannot keep up with agents that act faster than your reviewers.
Agents plan, call tools, and delegate to other agents with little supervision. Eighty-four percent of teams fail agent compliance audits using ordinary logs. We describe per-action zero-knowledge proof chains that bind identity, policy scope, and authorisation without exposing prompts or weights.
Summary
Agents plan, call tools, and delegate to other agents with little supervision. Eighty-four percent of teams fail agent compliance audits using ordinary logs. We describe per-action zero-knowledge proof chains that bind identity, policy scope, and authorisation without exposing prompts or weights.
Download the full PDF for technical detail, diagrams, and reproduction steps. Public sandbox: affix-io.com/sandbox.
Related reading
Frequently asked questions
Why do logs fail for agentic AI?
Agents mutate state quickly, share credentials, and get updated mid-session. Logs are editable and hard to replay after the fact.
What is an agent proof chain?
A linked sequence of compact proofs, one per action, each signed and Merkle-anchored so auditors verify behaviour without vendor access.
Which regulations apply?
EU AI Act Articles 11 to 15, NIST AI agent guidance, ISO 42001, and OWASP Agentic Top 10 controls map directly to proof artefacts.