Regulation · EU AI Act
Logs and human oversight, at the moment of action.
Regulation (EU) 2024/1689, read for AI agents that act: what the Act asks, what elan1 helps you evidence, and what stays with you and your counsel.
The EU AI Act asks high-risk AI systems to record events automatically (Article 12) and to be designed for effective human oversight (Article 14), and deployers must keep those logs (Article 26). elan1 helps by recording every agent action in a proof record and holding consequential writes for a named person. It does not make a system compliant alone.
elan1 does not certify you or itself against EU AI Act.
01What it asks
A risk-based law, with the heaviest duties on high-risk systems.
Cited to the official text. Where the Act sets a date, we show it; where amendments are proposed, we say so.
- Risk-basedSome practices are prohibited; high-risk systems carry the heaviest duties. Annex III includes creditworthiness assessment of individuals and life and health insurance risk assessment and pricing.
- ProvidersProviders of high-risk systems: risk management (Art. 9), data governance (Art. 10), technical documentation (Art. 11), record-keeping (Art. 12), information for deployers (Art. 13), human oversight (Art. 14), accuracy, robustness and cybersecurity (Art. 15).
- DeployersDeployers (Art. 26): assign human oversight to competent people; keep the logs under their control for at least six months unless other law says otherwise.
- TimingIn force 1 Aug 2024; duties phase in from 2 Feb 2025 to 2 Aug 2027. Amendments have been proposed — check the dates in force.
| Date | What applies |
|---|---|
| 1 Aug 2024 | The Act enters into force. |
| 2 Feb 2025 | Prohibited practices and AI literacy duties apply. |
| 2 Aug 2025 | General-purpose AI model duties and governance provisions apply. |
| 2 Aug 2026 | Most remaining provisions apply, including Annex III high-risk systems, unless amended. |
| 2 Aug 2027 | High-risk systems that are safety components of products under Annex I. |
Source: Regulation (EU) 2024/1689 (Artificial Intelligence Act), Official Journal of the European Union, 12 July 2024, Art. 9–15, 26 and 113, Annex III. Amendments have been proposed; check the dates in force.
02How elan1 helps
Where elan1 acts, and the evidence it leaves.
Four rows where a governed action produces evidence. The last column says whether it happens at the moment of action, and in which mode.
| What the Act asks | elan1 mechanism | Evidence | At the moment of action? |
|---|---|---|---|
| Record-keeping (Art. 12) | Proof record for every governed action, refusals included | Actor F09–F10, rules F06–F08, record F12–F13, time F16 | Yes, both modes |
| Human oversight (Art. 14, 26) | Named approver, not the requester; autonomy ceiling A0–A5; fail toward a human | Approvals; autonomy F11 | Mode A enforced; Mode B recorded |
| Keep logs (Art. 26) | Append-only chain; no UPDATE or DELETE on evidence rows | The chain | — |
| Risk management (Art. 9) — supports only | Pre-gates G1–G9, guardrail, versioned rules | Declaration version | Mode A enforced |
“Enforced” applies only in Mode A, where elan1 holds the write credential. In Mode B every write intent is evaluated and recorded; we call that attested advisory, never blocked.
03Not covered
What elan1 helps with, and what it does not.
Stated at equal weight, so nobody reads a gap as a feature.
Helps you produce evidence for
- Record-keeping for every governed action, refusals included (Art. 12)
- Human oversight by a named approver who is not the requester (Art. 14, 26)
- Keeping logs in an append-only chain (Art. 26)
- Supporting risk controls with pre-gates, a guardrail and versioned rules (Art. 9, supports only)
Does not cover
- Classifying your system
- Conformity assessment, CE marking, registration
- Your quality management system
- Model documentation, training data and accuracy testing
- Transparency notices
- General-purpose model duties
- Fundamental rights impact assessments
elan1 runs no EU residency domain; new regions open only on a signed contract.
04Regulation mapper
Clause by clause, to the rung that acts.
Select a row to see which rung of the Right Path it maps to. Switch frameworks to compare; nothing here is a certification.
EU AI Act: Record-keeping (Art. 12) maps to rung 8 PROOF.
| Requirement | elan1 control | Evidence produced | Not covered | Enforced at action time? |
|---|---|---|---|---|
| Rung 8 PROOF: a proof record for every governed action, refusals included | Actor F09–F10, rules F06–F08, record F12–F13, time F16 | Logging inside your model or application outside the write | Recorded at the action, both modes | |
| Rung 6 AUTONOMY: named approver, not the requester; autonomy ceiling A0–A5; fail toward a human | Approvals; autonomy F11 | Choosing and training the people you assign to oversight | Mode A enforced; Mode B recorded | |
| Append-only chain; no UPDATE or DELETE on evidence rows | The chain | Your retention schedule, and other law that sets it | After the action | |
| Rungs 3–5: pre-gates G1–G9, versioned rules, guardrail | Declaration version F07 | The risk management system itself | Yes, Mode A only | |
| None | None | Model documentation and instructions for use are yours | No. Outside elan1 | |
| None | None | Not covered. elan1 does not certify you or itself | No. Outside elan1 |
Where this is weak: Mode B records the write; it cannot stop it. “Enforced” applies only in Mode A, where elan1 holds the write credential.
Clause references summarise the official text. Not legal advice. elan1 holds no certificate against any framework. Source: Regulation (EU) 2024/1689, Official Journal of the European Union, 12 July 2024.
Read the full page: EU AI Act: what it asks and how elan1 helps
05Worked example
One rule, its clauses, and the record it writes.
A lender's rule for credit-limit increases, written once as a declaration, serves three articles and writes one record per decision.
1 · The rule, in the declaration
# decl_credit_limit_change · v3 (illustrative) write_type: credit_limit.increase pre_gates: [G3_PERMISSIONS] when: amount_inr > 500000 autonomy_ceiling: A1 approver: role:credit_officer approver_not_requester: true on_no_answer: refuse after 24h record: always # refusals included
Credit-limit increases above ₹5,00,000 need a named credit officer who is not the requester. No answer within 24 hours is a refusal. Every outcome writes a record.
2 · The clauses it serves
- Annex IIICreditworthiness assessment of individuals is listed as high-risk.
- Art. 14A competent person can oversee and stop the action.
- Art. 26The deployer assigns oversight and keeps the logs.
- Art. 12Events are recorded automatically over the system's life.
3 · The proof record it writes
- F05credit_limit.increase
- F06decl_credit_limit_change
- F073
- F09usr_specimen_credit_officer
- F10human
- F11A1
- F12acct_specimen_2207
- F162027-08-02T11:05:44+05:30
Illustrative specimen. Not customer data.
06What an auditor sees
One record, each field mapped to its clause.
At L1 the chain recomputes, showing nothing was changed or removed; from L2 the record arrives signed and verifies offline.
- F01tenant_id
- tnt_specimen_01
- F02seq
- 1042
- F03prev_digest
- sha256:9f3c…e41a
- F04write_id
- wr_specimen_7c2
- F05write_type
- journal_entry.post
- F06declaration_id
- decl_month_end_close
- F07declaration_version
- 7
- F08policy_version
- 2027.06.1
- F09actor_id
- usr_specimen_approver
- F10actor_kind
- human
- F11autonomy_level
- A2
- F12record_id
- je_specimen_5531
- F13record_version
- 2
- F14payload_hash
- sha256:51b0…7d9a
- F15lexicon_hash
- sha256:e2c4…0b16
- F16recorded_at
- 2027-07-14T10:42:07+05:30
Illustrative specimen. Not customer data.
Field notes · select a note to mark its fields
07FAQ
EU AI Act: questions we are asked.
Article 12 requires high-risk AI systems to allow events to be recorded automatically over the system's lifetime, so that situations that may create risk can be identified and the system's operation monitored. For an agent that acts, that means a record of each action: who or what acted, which rules applied, what changed and when. elan1 writes that record, a proof record, for every governed action, refusals included.
Article 14 asks that high-risk systems be designed so that people can oversee them effectively while in use: understand what the system does, decide not to use its output, and intervene or stop it. For an agent, that means a competent, named person sees a consequential action before it happens and can refuse it. elan1 holds writes above the autonomy ceiling for a named approver who is not the requester, and no answer in time is a refusal.
Not as a category. The Act classifies AI systems by use: an agent is high-risk when it is used in an area listed in Annex III, such as creditworthiness assessment of individuals or employment decisions, or as a safety component under Annex I. Whether your system is high-risk is a classification you make with your counsel.
No. elan1 helps you produce evidence for record-keeping and human oversight at the moment of a governed action. Classifying your system, conformity assessment, documentation and your quality management system stay with you and your counsel. elan1 does not certify you or itself.
No. A model may produce a write intent; it may never produce an allow. Rungs 3 to 8 of the Right Path cannot reach a model client, and evaluate() applies versioned rules only.
In Mode A, elan1 holds the write credential, so no write reaches your system of record without a gate decision and a proof record: that is enforced. In Mode B, reached over MCP or HTTP without the credential, every write intent is evaluated against your rules and recorded, but elan1 cannot stop the write. We state each integration's mode in writing before you rely on it.
At least six months, unless other EU or national law says otherwise (Article 26(6)). The duty covers the logs the high-risk system generates automatically, to the extent they are under the deployer's control. elan1 evidence rows are never updated or deleted; how long you keep them is a retention decision for you and your counsel.
From L2, a record arrives signed and verifies offline. At L1 the chain is tamper-evident to you: recomputing it shows whether anything was changed or removed.
Not today. elan1 runs one residency domain, "in" (AWS ap-south-1, Mumbai). An EU residency domain opens only on a signed contract.
Last reviewed
Reviewed against the official text. Not legal advice; your counsel decides how the law applies to you.
| Date | Change |
|---|---|
| 6 Oct 2026 | First version. Mapped Art. 9, 12, 14 and 26 to elan1 mechanisms; listed what is not covered; noted that amendments have been proposed. |
| 10 Oct 2026 | Reviewed at launch: wording updated now that elan1 Govern, Work, Decide and Run are available. Clause references unchanged. |
// NEXT STEP
Find out where you stand, in five minutes.
Twelve questions. A plain result. Nothing is stored or sent, and no email is needed.