The EU AI Act Rules That Went Live Today
Today, August 2, a real chunk of the EU AI Act stopped being theoretical. But if you followed the coverage, you probably got the wrong impression, because the story changed halfway through the year. A set of amendments quietly pushed the scariest obligations years down the road, while leaving a different, broader set live as of today. Getting that distinction right matters, because it determines whether you have a small housekeeping task this week or nothing to do at all, and either way it is not the compliance emergency the headlines imply.
What actually turned on today
Two things became enforceable. The transparency obligations, meaning people must be told when they are dealing with an AI system, and AI-generated or manipulated content like deepfakes and synthetic media must be labeled. And the enforcement powers over providers of general-purpose AI models, so the EU can now request documentation, test models, demand fixes, and impose fines rather than merely publish rules. Of the two, the transparency duties are what touch ordinary businesses, because almost everyone now has a chatbot or publishes AI-assisted content, while few operate a frontier model.
The part the headlines got wrong
For months, August 2 was billed as the day the AI Act reached "full applicability," including the strict rules for high-risk AI (hiring, credit, essential services, and the like). Then, in mid-2026, the Digital Omnibus amendments moved those deadlines.
| Obligation | When it applies |
|---|---|
| Transparency (chatbot + content labeling) | Live now (August 2, 2026) |
| General-purpose AI enforcement and fines | Live now (August 2, 2026) |
| High-risk standalone systems | Deferred to December 2, 2027 |
| AI embedded in regulated products | Deferred to August 2, 2028 |
So the heavy lifting most businesses feared is genuinely years away. What is live is the lighter, broader transparency layer. If you read our earlier note that the Act reached full applicability today, treat this as the update: the Digital Omnibus changed that, and only the transparency and general-purpose rules landed on schedule.
Whether it reaches you, and what to do
Like Europe's privacy law, these rules can apply based on where your AI is used, not only where you sit. If EU users interact with your chatbot, or you publish AI-generated content to EU audiences, you can be in scope from a Canadian office. Many small businesses with no EU footprint are not. Where it does apply, the actions are modest: add a clear note that users are talking to AI, label AI-generated or meaningfully AI-edited media, and support machine-readable marking for generative content. There is even a grace period, generative systems already on the market before today have until December 2, 2026 to meet the marking duty.
Where this leaves you
Do the ten-minute check: do EU audiences meet your AI, through a chatbot or AI-generated content? If not, note that you are likely out of scope and keep your attention on Canada's own evolving rules. If yes, add the disclosures, start labeling, and diarize the December marking deadline. Above all, do not let "high-risk" headlines panic you into an overhaul, that part is years out. What went live today rewards the same thing good businesses already value: being honest with people about when they are dealing with AI.
Frequently Asked Questions
What exactly became enforceable on August 2?
Two things. First, the EU AI Act’s transparency obligations (Article 50): people must be told when they are interacting with an AI system, and AI-generated or manipulated content such as deepfakes and synthetic media must be labeled. Second, the enforcement and fining powers over providers of general-purpose AI models. In plain terms, from this date the EU can actually request documentation, test models, demand corrections, and levy fines, rather than the rules simply existing on paper. The transparency duties are the part that touches the widest range of ordinary businesses.
Did the high-risk rules also start today?
No, and this is the point most coverage gets muddled. A set of amendments called the Digital Omnibus, finalized in mid-2026, pushed the strict obligations for high-risk AI back: to December 2, 2027 for standalone high-risk systems, and August 2, 2028 for AI embedded in regulated products. What did not move is the transparency obligations and the general-purpose AI enforcement, both of which are live now. So if you heard the AI Act reached "full applicability" today, that is outdated. The heavy compliance is deferred; the labeling rules are here.
I am a Canadian business. Does this reach me?
It can, the same way EU privacy law reaches beyond Europe. The transparency rules apply based on where your AI is used, not only where you are based. If you operate an AI chatbot that EU users interact with, or you publish AI-generated content to EU audiences, you can fall within scope from an office in Canada. Many small Canadian businesses with no EU footprint are not affected. But if you serve EU customers or your AI-driven output reaches people there, the labeling duties are worth taking seriously now.
What do the transparency rules actually require me to do?
Practically, three things where they apply. Tell people when they are talking to an AI rather than a human, a clear note on your chatbot covers this. Label AI-generated or meaningfully AI-edited content, including synthetic images, audio, and video, so audiences know it is not a genuine recording. And, for generative AI systems, support machine-readable marking of AI content. There is a short grace period: generative systems already on the market before August 2 have until December 2, 2026 to meet the machine-readable marking duty. None of this requires a legal department; it requires deciding to be upfront.
What should a Canadian business do this week?
Run a quick check. Do EU users interact with an AI chatbot on your site, or do you publish AI-generated content to EU audiences? If no, note that you are likely out of scope and focus on Canada’s own evolving rules. If yes, add a plain disclosure that users are talking to AI, start labeling AI-generated media aimed at EU audiences, and note the December marking deadline for existing generative systems. And do not overreact to the "high-risk" headlines, those obligations are years away. The move today is honesty about AI use, not a compliance overhaul.
Know exactly which AI rules apply to you
We help Canadian businesses cut through AI regulation, confirm what actually applies, and put right-sized disclosures and practices in place.
Related Articles
OSFI E-23: Canada's Model Risk Rules Meet AI
The EU AI Act Deadline: Does It Reach Your Business?
Your Team Is Doing Business on WhatsApp
AI consultants with 100+ custom GPT builds and automation projects for 50+ Canadian businesses across 20+ industries. Based in Markham, Ontario. PIPEDA-compliant solutions.