Eigma
EN 中 DE
The EU AI Act in Everyday Life Open Eigma
Contents The 2026 baseline

Explanations and enforcement

The 2026 baseline

The practical judgement

On 12 September 2026, major prohibitions, specified transparency duties, some general-purpose model obligations and complaints are enforceable, but most high-risk duties are not.

Article 113, amended in July 2026, made 2 August 2026 the general application date. National authorities and the AI Office began enforcing applicable provisions then.1 Under most Article 5 prohibitions, providers, deployers and other operators have been barred from placing, commissioning or using covered practices since 2 February 2025. National market-surveillance authorities, or the AI Office within its remit, have enforced those bans since 2 August 2026.2

This figure separates the current baseline from deferred safeguards.

<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 590 400" width="590" height="400" font-family="system-ui, sans-serif" font-size="14">
<title>AI Act provisions applying in the 2026 baseline</title>
<rect x="20" y="18" width="550" height="54" rx="8" fill="none" stroke="currentColor" stroke-width="2"/>
<text x="295" y="50" text-anchor="middle" fill="currentColor" font-size="17">AI Act field on 12 September 2026</text>
<line x1="295" y1="72" x2="295" y2="108" stroke="currentColor" stroke-width="2"/>
<polygon points="295,108 288,96 302,96" fill="currentColor"/>
<rect x="20" y="110" width="265" height="208" rx="8" fill="none" stroke="currentColor" stroke-width="2"/>
<text x="152" y="140" text-anchor="middle" fill="currentColor" font-size="16">Applicable baseline</text>
<text x="42" y="176" fill="currentColor">Most prohibited practices</text>
<text x="42" y="210" fill="currentColor">Specified transparency duties</text>
<text x="42" y="238" fill="currentColor">General-purpose provider duties</text>
<text x="42" y="260" fill="currentColor">Models placed from 2 Aug 2025</text>
<text x="42" y="282" fill="currentColor">Earlier models due 2 Aug 2027</text>
<text x="42" y="304" fill="currentColor">Article 85 complaint route</text>
<rect x="305" y="110" width="265" height="208" rx="8" fill="none" stroke="currentColor" stroke-width="2" stroke-dasharray="8 6"/>
<text x="437" y="140" text-anchor="middle" fill="currentColor" font-size="16">Deferred safeguards</text>
<text x="327" y="176" fill="currentColor">High-risk risk controls</text>
<text x="327" y="210" fill="currentColor">Data and technical records</text>
<text x="327" y="244" fill="currentColor">Human-oversight duties</text>
<text x="327" y="278" fill="currentColor">High-risk deployer duties</text>
<rect x="75" y="344" width="440" height="40" rx="8" fill="none" stroke="currentColor" stroke-width="2"/>
<text x="295" y="369" text-anchor="middle" fill="currentColor">Article 86 is applicable, but its current scope is disputed</text>
</svg>

Figure: The 2026 baseline combines enforceable rules with a large deferred high-risk regime.

Article 50 places specified transparency duties on providers and deployers from 2 August 2026. National market-surveillance authorities principally enforce them, with the AI Office covering its Article 75 remit and the European Data Protection Supervisor covering EU institutions.3 Chapter V has bound providers of general-purpose AI models placed on the market from 2 August 2025; providers of models placed earlier have until 2 August 2027. Under Article 88, the Commission and AI Office gained exclusive enforcement power on 2 August 2026.4

Article 85 also operates from 2 August 2026. The relevant market-surveillance authority must handle a complaint about a suspected infringement under its procedures and enforces whichever underlying duty is already applicable.5 Article 86 was not expressly deferred, so its explanation duty falls on deployers from that date and is enforced by national market-surveillance authorities or the AI Office within its remit. Its practical reach remains disputed because the associated Annex III classification rule was deferred.6

The core high-risk rules, including risk controls, data governance, records, human oversight and deployer duties, are absent from this baseline. Articles 8 to 26 bind providers and deployers from 2 December 2027 for Annex III systems and 2 August 2028 for product-embedded Annex I systems. Market-surveillance authorities or the AI Office within its remit enforce those duties only from the relevant later date.5

For action now, Article 85 complaints follow national procedures, with no EU-wide response deadline or guaranteed personal remedy.5 The AI Office tool applies only within its remit. Article 86 sets no prescribed request format, response deadline or express appeal mechanism.5 If personal data are involved, the established GDPR controller and data-protection authority routes may be more useful. The national AI authority list was still pending or blank in eighteen Member States.7

Boundary. General application does not mean every AI Act safeguard is active. New Article 5 bans concerning non-consensual intimate material and child sexual abuse material bind relevant providers and deployers only from 2 December 2026, when national market-surveillance authorities or the AI Office within its remit begin enforcing them.2

References

  1. [1]Regulation (EU) 2024/1689, consolidated text
  2. [2]Commission overview of AI Act enforcement
  3. [3]Commission FAQ on Article 50 transparency
  4. [4]Commission guidelines for general-purpose AI providers
  5. [5]Regulation (EU) 2024/1689, consolidated text
  6. [6]Open Legal Core analysis of Article 86
  7. [7]Commission list of AI Act market-surveillance authorities
Quizzes

  1. Which description best matches the AI Act on 12 September 2026?

    • Some major rules apply, while core high-risk duties remain deferred
    • Every operator safeguard already applies uniformly to all AI systems, regardless of risk category
    • Only voluntary guidance applies until the later high-risk dates

    The baseline includes prohibitions, transparency, model-provider obligations and complaints, but most high-risk controls start in 2027 or 2028.

  2. In the 2026 baseline, the core high-risk oversight and deployer duties are ____.

    • not yet generally applicable
    • already applicable in full
    • replaced by voluntary promises

    The amended schedule defers these core duties according to whether the system falls under Annex III or Annex I.

  3. A person may use the AI Act complaint route in 2026 even though many high-risk operator duties remain deferred.

    • True
    • False

    Article 85 applies from the general application date, but authorities can enforce only underlying duties whose own application dates have arrived.

Comments

No comments yet. Start the conversation.