official timeline checked 10 July 2026

EU AI Act Compliance
for AI Agents

AMC maps captured agent evidence to relevant obligations and shows what remains unsupported. It does not turn a maturity level or signed artifact into legal certification.

Why Self-Reported Documentation Won't Work

Documentation of intent is not the same as evidence that a control operated. AMC can capture agent behavior through its gateway, sign evidence events, and link mappings back to the underlying receipts.

Trust contract: a VALID artifact proves integrity, not evidence sufficiency. Only evidence readiness READY is claim-eligible. AMC does not certify legal compliance; qualified counsel must determine which obligations apply.

Maturity Is Not Legal Status

L0–L1

EARLY

Major control and evidence gaps remain.

DEVELOPING

Managed processes exist but evidence gaps remain.

DEFINED

Repeatable, measurable controls. Still requires READY evidence and obligation-specific review.

L4–L5

MANAGED

Stronger assurance and continuous verification, not automatic compliance.

Evidence Mapping, Not Certification

✓
Article 9 — Risk Management
Relevant assurance receipts can support risk identification and mitigation review. Applicability and sufficiency remain system-specific.
✓
Article 12 — Record Keeping
AMC hash-chains captured or ingested evidence events with Ed25519 signatures. Uncaptured activity is not represented in the ledger.
✓
Article 13 — Transparency
Reports expose L0–L5 scores, evidence references, and unsupported states. Reviewers can verify AMC artifacts offline with amc verify all --json.
✓
Article 14 — Human Oversight
Oversight scoring and signed action policies can provide evidence about approval boundaries; they do not establish legal conformity by themselves.
✓
Article 17 — Quality Management
Fleet consistency, policy canaries, monitoring, and drift alerts can support an evidence review when those controls are configured and operating.

Timeline

Feb 2025

AI Literacy obligations (Art. 4) took effect

Aug 2025

Prohibited AI practices ban entered into force (Art. 5)

Aug 2026

Article 50 transparency obligations apply for relevant interactive and generated-content systems

2 Dec 2027

Annex III high-risk rules apply under the current AI Omnibus political agreement timeline

2 Aug 2028

High-risk rules apply to systems integrated into regulated products

Source: European Commission high-risk guidance. Check the official source for later legal changes.

Get Started in 5 Minutes

No account needed. No API key. Free and open source.

$ curl -fsSL https://agentmaturity.co/install.sh | sh

$ mkdir my-agent && cd my-agent

$ amc init

$ amc compliance report --framework EU_AI_ACT

→ L2 (Developing) — mapping gaps remain; not a legal certification

$ amc audit binder create --framework eu-ai-act

→ Compliance evidence exported to .amc/audit/

$ amc compliance report --framework eu-ai-act

→ Gap analysis with per-article remediation steps

Build evidence before you need it.

Start with an honest baseline, capture real execution evidence, and keep legal review separate.

Get Started →

← Back to AMC Home · Full Documentation

EU AI Act Compliance for AI Agents