Problem Deploy Demo Security Pricing GitHub
●  Logs are not evidence
EU AI Act · Article 12 · Record-keeping

When the regulator asks,
what do you show?

Banks and fintechs now let AI agents approve loans, flag fraud, and screen applicants. When a regulator, a court, or an auditor asks you to prove what the agent decided, and that the record was not changed afterward, most teams have nothing that survives the question. EU AI Act Article 12 is one reason to close that gap. An adverse-action lawsuit, a SOC 2 audit, or an OCC exam is a sooner one. The Article 12 obligations for Annex III systems are deferred to 2 December 2027 under the Digital Omnibus, adopted by the Council on 29 June 2026 and pending publication in the Official Journal. Either way it is runway to do it right, not a reason to wait.

● The deadline moved. The exposure did not.
NowDiscovery requests, SOC 2 audits, and adverse-action / fair-lending litigation already test whether you can prove a decision.
In forceGDPR Article 22 · accountability for solely-automated decisions affecting people. Already enforceable.
2 Dec 2027EU AI Act Article 12 (Annex III standalone systems): deferred to 2 December 2027 under the Digital Omnibus, adopted by the Council on 29 June 2026 and pending publication in the Official Journal. (Annex I embedded systems: 2 August 2028.)
The first time a decision is challenged is the real deadline. It does not wait for a regulation. View Article 12 text →
What the law actually requires

Article 12 in plain language

Article 12 requires high-risk AI systems to automatically record events (logs) over the lifetime of the system, for traceability of how the system functions. Article 19 then requires those logs to be kept for at least six months. That is the legal floor. The list below separates what the regulation literally says from what examination-grade evidence actually looks like. A log a regulator, a court, or an auditor cannot trust is not yet evidence.

Law stated in the regulation Best practice exceeds the text · what makes a log trustworthy
01
Automatic recording of events
Law Art. 12(1) · automatic recording over the system lifetime

Article 12(1) requires the system to record events automatically, over its lifetime. Not logs a person exports or assembles after the fact. For an AI agent making consequential decisions, the defensible reading is that a record is generated at the moment of each decision, by the system itself. Many teams already meet this with their existing logging.

What you need

  • SDK-level or infrastructure-level capture at decision time
  • No manual export steps in your evidence workflow
  • Timestamp generated at point of decision, not at point of export
02
Tamper-evident storage with verifiable integrity
Best practice Integrity · not literal Art. 12 text

Article 12 requires logging but does not, in its words, mandate cryptographic integrity. Yet a record that could have been edited after the fact is weak evidence to a regulator, a court, or an auditor. This is where evidence diverges from logging: the reader has to be able to confirm, independently, that no record changed after it was written. Most logging infrastructure cannot provide that proof.

What you need

  • Cryptographic signing of each artifact at time of creation
  • Write-once storage (WORM) preventing post-commit modification
  • Independent verifiability with no vendor cooperation required
  • Timestamping that proves the artifact existed at a specific time
03
Logging the full input to the decision
Law Art. 12(3) for biometric · Best practice otherwise

Article 12(3) explicitly requires input-data logging for remote biometric identification systems (Annex III, point 1(a)). For other high-risk systems the text does not mandate it. But if you cannot show the input a decision was actually based on, you cannot reconstruct or defend the decision. For agents that use retrieval, that means the assembled prompt and every document pulled in.

What you need

  • Full prompt snapshot at decision time, not a pointer to a template
  • For RAG systems: each retrieved document and its source
  • Cryptographic binding of the input to the decision artifact
  • Evidence that the captured input is the input that was actually used
04
Reconstructing multi-step and multi-agent workflows
Best practice Art. 12(2) traceability, applied to agents

Article 12(2) frames logging around traceability of the system's functioning. It does not spell out multi-agent chains. The Act predates today's agent architectures. But the principle is clear: if a decision spans several steps or hands off between agents, a single record at the end does not let anyone reconstruct what happened or why.

What you need

  • Hash-linked records from each step referencing the prior step's artifact
  • Capture of every tool call made during the decision workflow
  • Record of which agents were involved and in what sequence
  • Evidence the chain has not been truncated or selectively omitted
05
Retention and accessibility for examination
Law Art. 19 · logs kept at least six months

Retention lives in Article 19, not Article 12: logs must be kept for a period appropriate to the intended purpose, and at least six months, unless other law requires longer. Many teams already meet the retention bar; the harder part is that the records have to be usable: readable without your internal systems, and not dependent on a vendor still being around to interpret them.

What you need

  • Defined retention period (six-month legal minimum) with access controls
  • Standalone verification: a reader verifies without your infrastructure
  • Escrow arrangement if artifacts are held by a third-party vendor
  • Export format readable without proprietary tooling
REFOfficial text: Article 12: Record-keeping and Article 19: retention of logs. Article 12 applies to all high-risk AI systems (Annex III and Annex I). Annex III standalone high-risk obligations are deferred to 2 December 2027 (Annex I embedded systems to 2 August 2028) under the Digital Omnibus, adopted by the Council on 29 June 2026 and pending publication in the Official Journal. Article 50 transparency obligations remain applicable from 2 August 2026. The same evidence answers more than one reader. EU AI Act Art. 12, GDPR Art. 22, a SOC 2 control, and a litigation discovery request are largely asking for the same artifact. Items marked "best practice" exceed the literal text. Current as of June 2026. Not legal advice.
Self-assessment

Could you prove it today?

Check what your team can demonstrate right now, not what you plan to build. The first item is the Article 12 legal baseline; the rest are what turn a log into evidence that holds up when a decision is challenged.

This is a reality check, not a sales funnel. Several of these you can satisfy with infrastructure you already run. If you can check most of them with your own stack, you do not need a vendor. Only check what you could actually show a regulator, a court, or an auditor today. Fill it out with whoever owns model risk or AI governance and your engineering lead.

Our AI agents generate a log record automatically at every consequential decision, with no manual step required
Law · Art. 12(1) · often met with existing logging
We have a defined retention policy (≥ 6 months) with access controls, and records cannot be casually deleted
Law · Art. 19 · most teams can meet this without a vendor
Every record is cryptographically signed at time of creation and we can prove no record has been altered
Best practice · integrity
Records are in write-once (WORM) storage that prevents modification even by an administrator
Best practice · immutability
We capture the full input (assembled prompt and retrieved context) for every logged decision
Law for biometric (Art. 12(3)) · best practice otherwise
We can reconstruct the full decision chain for multi-step agent workflows, not just the final output
Best practice · Art. 12(2) traceability
A third party can verify any decision record using only a public key, with no access to our systems or vendor required
Best practice · independent verifiability
We have trusted timestamps (e.g. RFC 3161) proving each record existed at a specific time
Best practice · timestamping
Evidence Readiness Score0 / 8
Check items above to see where you stand.
Next step

Book a 15-minute evidence-readiness call

We will look at what you have, tell you what would survive a regulator, a court, or an auditor, and where the gaps are. If your existing stack already covers it, we will tell you that. No pitch.

Book 15-minute call →

Forward to whoever owns model risk, AI governance, or financial-crime engineering. They will know if this is real for your team.
URL: plainreal.com/article12

This page is a practical guide for compliance and engineering teams, current as of June 2026. It is not legal advice. Confirm the current status of the Digital Omnibus deferral with your counsel before making compliance determinations. Maintained by AgentOS Technologies, Inc.