The EU AI Act Timeline: Key Dates and Deadlines for Publishers

The EU AI Act is the most consequential piece of technology regulation to hit European publishers in a generation, yet its phased structure makes it genuinely difficult to track. Different obligations activate on different dates, some deadlines were extended by the May 2026 AI Omnibus package, and the penalties for non-compliance are severe. What follows is a plain-language timeline of every date that matters, written for editors and publishing executives who need clarity without the legal jargon.

The Foundation: Entry into Force

The AI Act entered into force on 1 August 2024. That date did not immediately impose any operational duties on publishers, but it started the clock on everything that follows. From that moment, the European Commission and member states began building the governance infrastructure the law requires, and organisations were expected to begin assessing their exposure. If your newsroom was still treating the Act as a distant concern after August 2024, that window has closed.

2 February 2025: The First Real Deadline

Six months after entry into force, two categories of obligation became enforceable. First, the prohibition on unacceptable-risk AI practices took effect. These are the uses the Act bans outright: subliminal manipulation, social scoring, certain biometric categorisation systems, and similar practices that sit far outside the editorial toolkit. No publisher should have been close to these lines, but the date matters because it confirmed the Act was live and coercive.

Second, and more practically relevant, AI literacy duties activated on the same date. Article 4 of the Act requires deployers and providers to ensure their staff possess a sufficient level of AI literacy. For newsrooms, that translates into a concrete obligation to train journalists, editors, and technologists on how the AI tools they use actually work, what their limitations are, and how to identify when outputs may be unreliable. We cover the practical side of that obligation in our guide to AI writing assistant policies for editors.

2 August 2025: GPAI Model Obligations

One year after entry into force, obligations targeting general-purpose AI model providers came into effect. These rules apply primarily to the companies building large foundation models, not to the newsrooms licensing or deploying them. However, publishers who have built or fine-tuned their own models on proprietary data, or who distribute AI-generated outputs at scale, should assess carefully where they sit in the supply chain. The distinction between a deployer and a provider is not always obvious, and regulators are likely to examine it closely.

2 August 2026: The Main Event

The majority of the Act's obligations, including Article 50 transparency requirements, apply from 2 August 2026. Article 50 is the clause that matters most for day-to-day editorial operations. It requires that AI-generated or AI-manipulated content, particularly synthetic audio, video, and images, be labelled in a machine-readable format so that it can be detected and disclosed to audiences. This is not a voluntary best-practice standard. It is a legal requirement with enforcement teeth.

The transparency debate inside newsrooms is already well underway. Our analysis of how publishers are approaching AI content labelling shows that editorial cultures vary considerably on where to draw the line between AI-assisted and AI-generated work, but the Act cuts through some of that ambiguity by focusing on the output rather than the process.

One important grace period sits within this phase: generative systems already on the market before 2 August 2026 have until 2 December 2026 to meet the machine-readable marking requirement under Article 50(2). That four-month window is intended to give existing deployments time to retrofit compliance, not to delay planning.

For a detailed account of what changed between the Act's original text and its current state, see our 2026 status update for publishers.

The Omnibus Delays: 2027 and 2028

The May 2026 AI Omnibus package introduced significant postponements for high-risk AI obligations. Annex III high-risk systems, a category covering AI used in areas such as employment, education, and critical infrastructure, now face a compliance deadline of 2 December 2027, pushed back from the original 2026 date. Annex I high-risk rules, which cross-reference existing sectoral safety legislation, move to 2 August 2028.

For most general publishers, the Annex III and Annex I categories will be peripheral rather than central. However, broadcasters using AI in hiring or audience-profiling decisions, and investigative teams deploying AI-assisted analysis tools in sensitive contexts, should map their systems against both annexes before assuming they are unaffected.

Penalties: Why the Dates Are Not Negotiable

The Act's enforcement regime is designed to ensure compliance is not treated as optional. Fines for prohibited-practice violations reach up to 35 million euros or 7 percent of global annual turnover, whichever is higher. Lower tiers apply to other infringements, but the scale of the top-line penalty signals that the European legislature intended the Act to be taken as seriously as the GDPR. Publishers with international parent companies should note that the turnover calculation is global, not limited to EU revenues.

Framing AI compliance purely as a legal risk undervalues it, though. The deeper argument, which we make in our editorial guidelines on synthetic media ethics, is that audience trust depends on publishers getting ahead of these standards rather than being pulled toward them by enforcement.

A Consolidated Timeline at a Glance

• 1 August 2024: AI Act enters into force.
• 2 February 2025: Prohibited practices banned; AI literacy duties apply.
• 2 August 2025: GPAI model obligations apply.
• 2 August 2026: Article 50 transparency and most core obligations apply.
• 2 December 2026: Grace period ends for Article 50(2) machine-readable marking on pre-existing generative systems.
• 2 December 2027: Annex III high-risk obligations apply (postponed by the Omnibus).
• 2 August 2028: Annex I high-risk rules apply (postponed by the Omnibus).

The practical implication for newsrooms is that the 2 August 2026 date should be treated as the operational deadline. Everything else is either already past or relates to categories that require specialist legal mapping. For guidance on the content-authenticity infrastructure that underpins Article 50 compliance, our publisher's guide to content authenticity is the right starting point.

Sources

• Regulation (EU) 2024/1689 of the European Parliament and of the Council (the EU AI Act), Official Journal of the European Union, published 12 July 2024.
• European Commission, AI Omnibus package, May 2026.
• European Parliament legislative observatory, EU AI Act timeline records.