Agent Compliance:
Be Audit-Ready

EU AI Act obligations are here. Can you prove, years from now, what your AI decided today?

EU AI Act Timeline

Feb 2025 Article 5 — prohibited practices took effect. Social scoring, manipulation exploiting vulnerability, and other banned uses are now unlawful.
Aug 2026 High-risk AI obligations apply — including governance, transparency, and human oversight duties for high-risk systems.
Article 19 Automatic log retention — high-risk AI systems must automatically record events for at least 6 months, and retain them for the system's lifetime.

EU AI Act fines are tiered by violation type:

Article 5 (prohibited practices): up to €35 million or 7% of worldwide annual turnover.
Other obligations (incl. logging): up to €15 million or 3% of worldwide annual turnover.

The requirement to prove how a decision was made doesn't go away when the quarter ends. The evidence has to survive the audit.

Tamper-evident audit chain — every action is Ed25519-signed and Merkle-linked, so logs can't be silently altered.
Offline verification CLI — prove the integrity of the audit trail on your own infrastructure, without shipping data to a third party.
Accountable identity — every agent gets a per-agent identity (badge); every action binds to it.
Human review lane — a 4-level decision tree routes only a small share of actions to human review, keeping high-risk decisions supervised.
Retention-ready logging — automatic event logging aligned with the 6-month retention requirement.
Chinese national cryptography — SM2/SM3/SM4-GCM support (SM family: GB/T 32918 SM2, GB/T 32905 SM3, GB/T 32907 SM4) for deployments that require domestic cryptographic standards.

✓ Record every AI decision with a tamper-evident trail
✓ Retain automatic logs for the required period (6 months for high-risk AI systems)
✓ Bind every action to an auditable, accountable identity
✓ Keep a human review lane for high-risk actions
✓ Be able to reproduce, offline, how a decision was made