August 2 Has Passed. Here's What Actually Changed — and What Didn't.

August 2, 2026 was the EU AI Act's general applicability date, and with it Article 50's transparency obligations began applying. The duties split by role, and that split is the whole story for agencies. Deployer-side duties — the ones relevant to an agency running AI in client work — are the narrower set: disclosing AI-generated or AI-altered content (including deepfakes) to the audiences that encounter it. Those norms are no longer upcoming; they are in effect.

Just as important is what did not change on August 2. Article 50(2)'s machine-readable marking requirement is a provider obligation — it sits with the companies that build and supply AI systems — and the EU's Digital Omnibus moved that marking duty to December 2, 2026 for systems on the market before August 2, 2026. And in the United States, California's SB 942 continues to target covered providers above one million monthly active users, not every agency that uses AI tools. If a headline this week implied that every business using AI crossed a legal cliff on Saturday, it overstated the situation.

For most agencies, the practical pressure was never a regulator arriving on August 3. It comes downstream: EU-headquartered clients whose procurement and compliance teams now ask 'how do you disclose AI in the work you deliver?' — and platforms and vendor questionnaires asking the same thing in writing. Before August 2, a thin answer read as early. Now it reads as a gap.

The calm response is structure, not panic: know which of your public-facing deliverables are AI-generated or AI-assisted, disclose where audiences encounter them, keep a labeling step in the campaign process, and hold an evidence trail you could produce on request. That is a process question more than a legal one — and it is auditable, fixable, and worth having answered before the next renewal conversation asks it for you.