The EU AI Act Just Hit Its Biggest Deadline. Is Your AI Stack Ready?

On 2 August 2026 the EU AI Act reached its general application date. A last-minute amendment pushed the high-risk rules back, but transparency duties, enforcement powers over model providers and the rest of the Act still landed on schedule. If you run AI chatbots, AI-generated content or AI agents that touch EU users, here is what changed and what to do about it.

This article is general information, not legal advice. The AI Act is being implemented in stages and amended along the way, so check the current consolidated text and talk to counsel before you make compliance decisions.

The Short Version: What Applies Right Now

Article 113 of the AI Act says the regulation "shall apply from 2 August 2026," with a set of earlier and later exceptions (AI Act Explorer, Article 113). As of October 2026, the timeline looks like this:

• 2 February 2025: General provisions and prohibited practices (Chapters I and II), including the AI literacy duty in Article 4 (Article 4).
• 2 August 2025: Governance, the general-purpose AI (GPAI) model rules in Chapter V and the penalties chapter. Article 101, which covers fines on GPAI providers, was specifically excluded from that date (Article 113).
• 2 August 2026: Everything else unless it was postponed. That includes the Article 50 transparency obligations (Freshfields).
• 2 December 2026: End of the grace period for machine-readable marking of AI content, for systems placed on the market before 2 August 2026. Also the deadline for the new bans on nudifier and CSAM-generating systems (Freshfields).
• 2 December 2027: Stand-alone high-risk systems under Annex III. This was postponed from 2 August 2026 (Gibson Dunn).
• 2 August 2028: High-risk AI embedded in regulated products under Annex I (Gibson Dunn).

The Delay Everyone Was Waiting For: The Digital Omnibus

The high-risk regime was originally due on 2 August 2026. The EU's "Digital Omnibus on AI" moved it. Gibson Dunn reports that a provisional political agreement was reached on 6 May 2026 and confirmed by the Council on 13 May (Gibson Dunn, 27 May 2026). According to Freshfields, Parliament approved the final text on 16 June 2026, the Council approved it on 29 June, and it entered into force in July 2026 (Freshfields, 10 July 2026).

The changes most teams will notice:

• High-risk deadlines moved: 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems.
• Transparency did not move: Article 50 still applies from 2 August 2026. The only exception is a grace period to 2 December 2026 for watermark-style marking on systems already on the market.
• AI literacy was softened: Article 4 now requires providers and deployers to "take measures to support the development of AI literacy" of their staff. It no longer requires them to guarantee a particular level (Article 4; Gibson Dunn).
• New prohibitions: AI systems that generate non-consensual intimate imagery or child sexual abuse material are banned, with a transition period to 2 December 2026.

The high-risk delay matters if you use AI in an Annex III area. Those areas include employment, for example systems for "filtering applications" and evaluating candidates, plus biometrics, education, critical infrastructure and access to essential services (Annex III). You have more time, but the obligations themselves are unchanged.

Provider or Deployer? Figure Out Which Hat You Wear

Most AI Act duties depend on your role. A provider develops an AI system or GPAI model, or has one developed, and places it on the market under its own name. A deployer is anyone "using an AI system under its authority" outside purely personal use (Article 3).

A company that uses AI tools and agents internally is usually a deployer. If you put a branded chatbot or AI feature in front of customers, you may also be the provider of that system, even when a third-party model runs underneath it.

Article 50: The Transparency Duties That Just Switched On

Article 50 is the part of the Act that reaches the most ordinary businesses (Article 50):

• Chatbots must say they are AI. Providers have to make sure people are told "that they are interacting with an AI system" unless that is obvious from context.
• Synthetic content must be machine-detectable. Providers of systems that generate audio, images, video or text must mark the output "in a machine-readable format and detectable as artificially generated or manipulated."
• Deepfakes must be labelled. Deployers must disclose deepfakes. They must also disclose AI-generated text published to inform the public on matters of public interest, unless a human has reviewed it and someone holds editorial responsibility.
• Emotion recognition and biometric categorisation: Deployers must tell the people who are exposed to these systems.
• Timing: The information must be given "in a clear and distinguishable manner at the latest at the time of the first interaction or exposure."

To help with this, the European Commission published a voluntary Code of Practice on marking and labelling AI-generated content on 10 June 2026. Jones Day notes that the Code calls for "at least two layers of machine-readable marking where necessary, such as combinations of metadata, watermarks, or other technical measures." Deployers can use an EU icon or an alternative such as an audio disclaimer (Jones Day, June 2026).

GPAI Providers: Obligations Since 2025, Fines From 2026

If you only build on top of models from OpenAI, Anthropic, Google or xAI, the GPAI chapter mainly binds them, not you. It still affects you, because Article 53 requires GPAI providers to give downstream developers documentation on what their models can and can't do. Providers must also keep technical documentation, adopt a copyright policy and publish a summary of their training content (Article 53).

The new part is enforcement. Article 101 lets the Commission fine GPAI providers up to 3% of worldwide annual turnover or EUR 15 million, whichever is higher (Article 101). Article 113 kept Article 101 out of the 2025 start date, so those fines became available with general application in August 2026. Models placed on the market before 2 August 2025 have until 2 August 2027 to comply (Article 111).

What the penalties look like for everyone else

Article 99 sets fines of up to EUR 35 million or 7% of worldwide turnover for prohibited practices. Most other operator obligations carry fines of up to EUR 15 million or 3%, and supplying incorrect information to authorities up to EUR 7.5 million or 1%. For SMEs and start-ups, the lower of the two amounts applies (Article 99).

What This Means for Teams Running AI Agents

The AI Act has no separate category for "agents," so an agent is regulated by what it does. A practical checklist:

1. Inventory your AI touchpoints. List every chatbot, content generator, agent and automated workflow, and note whether it faces the public, whether it produces images, audio or video, and whether it touches an Annex III area such as hiring.
2. Add disclosures where agents talk to people. An agent that replies to customers on Messenger, WhatsApp or email in your name is the kind of system Article 50(1) covers.
3. Label what you publish. AI-written posts on matters of public interest, and any deepfake-style media, need a label unless a human review step with editorial responsibility applies.
4. Keep a human in the loop for consequential actions. Human review is the condition for the Article 50(4) text exemption. It is also a sensible control in its own right.
5. Document AI literacy efforts. The duty was softened, not removed. Short internal training and written usage guidelines are a reasonable baseline.
6. Plan for December 2027. If any workflow screens CVs or scores people, start your high-risk work now rather than in 2027.

How MultiverseAI Fits In

MultiverseAI doesn't make you compliant by itself. No tool does. It does give you controls that line up with the steps above. Approval gates stop an agent before it posts, sends or deletes anything, so a person reviews AI-generated content before it reaches the public. The visual Workflow Builder (55+ node types, plus an AI Workflow Builder that drafts a graph from one sentence) lets you see every step that runs on a schedule, webhook or chat trigger, which makes the inventory easier. Because multi-model chat covers OpenAI, Anthropic, Google and xAI models from one account, you can switch providers without rebuilding your stack as their documentation and terms change. All 50 integrations connect through your own accounts.

Ready to put human review in front of your AI? Create a free account (it comes with 100 free credits) or see how approval gates and workflows work.

Sources

1. EU AI Act Explorer, "Article 113: Entry into Force and Application." https://artificialintelligenceact.eu/article/113/
2. EU AI Act Explorer, "Article 50: Transparency Obligations for Providers and Deployers of Certain AI Systems." https://artificialintelligenceact.eu/article/50/
3. EU AI Act Explorer, "Article 4: AI Literacy." https://artificialintelligenceact.eu/article/4/
4. EU AI Act Explorer, "Article 3: Definitions." https://artificialintelligenceact.eu/article/3/
5. EU AI Act Explorer, "Article 53: Obligations for Providers of General-Purpose AI Models." https://artificialintelligenceact.eu/article/53/
6. EU AI Act Explorer, "Article 99: Penalties." https://artificialintelligenceact.eu/article/99/
7. EU AI Act Explorer, "Article 101: Fines for Providers of General-Purpose AI Models." https://artificialintelligenceact.eu/article/101/
8. EU AI Act Explorer, "Article 111: AI Systems Already Placed on the Market or Put into Service." https://artificialintelligenceact.eu/article/111/
9. EU AI Act Explorer, "Annex III: High-Risk AI Systems Referred to in Article 6(2)." https://artificialintelligenceact.eu/annex/3/
10. Gibson Dunn, "EU AI Act Omnibus Agreement — Postponed High-Risk Deadlines and Other Key Changes," 27 May 2026. https://www.gibsondunn.com/eu-ai-act-omnibus-agreement-postponed-high-risk-deadlines-and-other-key-changes/
11. Freshfields, "EU AI Act unpacked #34: The final Digital Omnibus on AI," 10 July 2026. https://www.freshfields.com/en/our-thinking/blogs/technology-quotient/eu-ai-act-unpacked-34-the-final-digital-omnibus-on-ai-key-amendments-to-the-a-102nber
12. Jones Day, "European Commission Publishes Final Code of Practice on Marking and Labelling AI-Generated Content," June 2026. https://www.jonesday.com/en/insights/2026/06/european-commission-publishes-final-code-of-practice-on-marking-and-labelling-aigenerated-content