The EU AI Act formally entered into force on 1 August 2024. It does not hit everyone on the same morning. Rules arrive in layers. For ordinary users and most companies, the date that finally feels "serious" is 2 August 2026: wider enforcement, and transparency duties that show up in products people actually open.
This post is a plain guide. Not a legal memo. If you need advice for a regulated product, talk to counsel. Here is what the calendar means in practice.
Timeline (why so many different dates)
- 1 August 2024 , AI Act formally enters into force.
- 2 February 2025 , bans on some AI practices start applying, along with definitions and AI literacy duties.
- 2 August 2025 , duties for providers of general-purpose AI models, plus parts of the governance setup.
- 2 August 2026 , transparency rules apply more broadly and wider enforcement of the Act kicks in. This is the public-facing hinge.
- 2 December 2026 , further bans, and some older systems that generate or alter synthetic content face a hard deadline to meet marking rules.
- 2 December 2027 , rules for some high-risk systems listed in Annex III start applying (examples: employment, education, critical infrastructure, migration).
- 2 August 2028 , later deadline for high-risk AI built into already regulated products (for example certain medical devices or toys).
Different articles quoting 2024, 2025, or 2026 are often all correct. They are pointing at different layers.
What changes for ordinary users around August 2026
You should start noticing clearer signals in everyday tools:
- A chatbot should make it clear you are talking to AI, not a silent human agent.
- Deepfakes and similar synthetic media should be labelled as such.
- Certain AI-generated or AI-modified content should carry marks that machines can detect, not only a tiny caption you can miss.
In practice that means fewer "gotcha" conversations with support bots, and a better chance to spot fake video or audio before you share it. It will not make every scam impossible. It raises the floor for honesty in mainstream EU-facing products.
What changes for companies
If you only embed a third-party chatbot, you still need to know how it presents itself to users and how outputs are marked when the rules say they must be. If you build or fine-tune models, 2025 already started GPAI duties; 2026 adds pressure to prove transparency in products that ship to EU users.
High-risk use (hiring filters, education scoring, infrastructure, border tools) gets its heavy chapter later, in 2027, with product-embedded cases stretching to 2028. Do not wait until those years to map where your tools sit. Inventory now: chat widgets, image tools, voice clones, CV screeners.
Also budget for process, not only engineering: who approves a new AI feature, who writes the user-facing notice, who keeps evidence when a regulator asks.
What this is not
It is not a ban on AI in the EU. It is not a guarantee that every app outside the EU will play nice. It is a staged rulebook that finally becomes visible to shoppers and workers when chatbots must say they are bots and deepfakes must stop pretending to be untouched reality.
Practical checklist for August 2026
- Users: look for clear AI notices; treat unmarked "human" support with suspicion.
- Creators: expect platforms to demand labels on synthetic media.
- Businesses: list every customer-facing AI; fix bot disclosure; plan marking for generated media; diary the later high-risk dates if you touch hiring, education, or regulated hardware.
Publish this beside the calendar above. The Act became law in 2024. For most people, it becomes felt in 2026.
Brak komentarzy:
Prześlij komentarz