European UnionAI Act, Article 57 (national AI regulatory sandboxes)
Regulation (EU) 2024/1689, Article 57
A citation is an address, not a summary. The first part names the law; what follows narrows it to the exact section, article or paragraph.
What follows is LexLint's own research summary of this law, not legal advice.
An AI governance rule binding government bodies.
As of 24 September 2026.
What it requires
- This is a Member State duty to establish infrastructure, not a duty on an AI provider or deployer. If you are a provider or prospective provider, you may apply to participate in your Member State's AI regulatory sandbox once established, but nothing in this Article requires you to.
- A provider or prospective provider that follows its agreed sandbox plan and the competent authority's guidance in good faith is not fined under this Regulation, or under other Union or national law actively supervised within the sandbox, for the conduct the sandbox covers.
If you get it wrong
Criminal exposureNo
Private right of actionNo
What it reaches
Obligation class
Governance
What it makes you log
Who may demand the log
Regulator, Public
Logging duty
National competent authorities, not providers or deployers, must submit annual reports on a sandbox's progress and results to the AI Office and the Board, and make the reports or an abstract public. This is a government-to-government reporting duty rather than a provider or deployer record-keeping duty.
- Kind
- Explicit
- As of
- 24 September 2026
- Provision
- Article 57(16)
- Trigger
- all_systems
Who checks it
Audit expectation
none
Also on the record
EEA status
- Status
- Pending
- Source link
- https://www.efta.int/eea-lex/32024r1689
What this law does
Drafted with AI from the cited sources under the direction of UnGovr staff. UnGovr holds editorial responsibility for this page.
Each Member State must ensure that its competent authorities establish at least one AI regulatory sandbox at national level, operational by 2 August 2027, a date the Digital Omnibus on AI (Regulation (EU) 2026/1744, in force since 27 July 2026) moved back one year from the original 2 August 2026 deadline; a Member State may instead participate in another Member State's sandbox if that gives equivalent national coverage.
A sandbox provides a controlled environment for developing, training, testing and validating innovative AI systems before they are placed on the market, under a plan agreed with the competent authority, and may include supervised real-world testing.
Article 57 creates no duty on a provider or a deployer: participation is voluntary, and a participating provider or prospective provider that follows its sandbox plan and the competent authority's guidance in good faith is shielded from an administrative fine under this Regulation, and from a fine under other Union or national law actively supervised in the sandbox, for the conduct the sandbox covers.
Competent authorities must give sandbox participants regulatory guidance, and, on request, written proof of activities and an exit report a provider may use to help demonstrate compliance.
When LexLint raises it
trains_modelsgenerates_contenthigh_risk_decisions
Read the law
official consolidated Official Journal text, EUR-Lex
Every line above is drawn from the primary source linked here, read on the date shown. This is a research summary, not legal advice.