EU AI Act Transparency Rules Take Effect Today Despite the 2027 Delay

As of August 2, 2026, the EU AI Act's transparency rules under Article 50 took effect, requiring AI systems serving EU users to disclose AI-generated content, while the European Commission gained enforcement powers over general-purpose AI providers, allowing fines up to 3% of global annual turnover or €15 million. The European Parliament's June 16, 2026 vote (423-57) and Council approval on June 29 delayed high-risk compliance to December 2, 2027, but left transparency and GPAI enforcement unchanged. Companies that treated the delay as a stand-down may face fines for non-compliance.

Everyone read the EU AI Act delay as a 16-month reprieve. It wasn't. As of today, August 2, 2026, AI systems serving EU users must disclose when content is AI-generated, and the European Commission can now fine general-purpose AI providers directly. You'd be forgiven for thinking the EU AI Act just went quiet for another year and a half. On June 16, 2026, the European Parliament voted 423 to 57, with 174 abstentions, to approve the Digital Omnibus on AI, and the Council of the EU signed off on June 29. The headline everywhere was simple: high-risk AI compliance pushed back to December 2, 2027. That's true, but it's also the part of the story that matters least today. Here's what didn't move. Article 50 of the AI Act, the transparency and disclosure rules, went live on schedule this morning. Any AI system that talks to a user, generates content, or manipulates images and audio now has to say so, whether or not it's classified as high-risk. Chatbots need to disclose they're bots. Deepfakes need labels. AI-generated text published as news needs a disclosure. None of that got delayed by the Omnibus. According to legal analysis from Gibson Dunn and White & Case, the transparency obligations were left essentially untouched by the negotiations that reshaped everything else. The other date that didn't move belongs to the model providers themselves. The Commission's enforcement powers over general-purpose AI systems, the category that covers OpenAI's GPT models, Anthropic's Claude, Google's Gemini, Meta's Llama, and xAI's Grok, also activate today. That means the Commission can now request internal documentation, run its own evaluations, and impose fines of up to 3% of global annual turnover or 15 million euros, whichever is higher, under Article 101 of the Act. Any GPAI model placed on the EU market since August 2025 has technically been under full obligations already, with no grace period. What changes today is that Brussels can actually enforce it. The Digital Omnibus really did rewrite the calendar for high-risk systems. Standalone high-risk uses under Annex III, think hiring, credit scoring, law enforcement, now have until December 2, 2027 to comply. AI embedded as a safety component in regulated products - medical devices, industrial machinery - gets until August 2, 2028. Genuine, substantial delays. Law firms from Morgan Lewis to Travers Smith have been advising clients on them for weeks. But high-risk classification and general-purpose transparency are two different chapters of the same law. A lot of companies seem to have conflated them anyway. One compliance-tracking publication put it bluntly: businesses treated the postponement headlines as a stand-down order, when the date that actually lands first, transparency and GPAI enforcement, never moved. If your product isn't high-risk under Annex III but it generates or manipulates content for EU users, the Omnibus vote changed nothing this week. Frankly, that's the trap. A 16-month delay is the kind of headline that gets skimmed once and filed away. A narrower carve-out that keeps a chunk of the law active is the kind of detail that gets missed, and missing it now carries a real fine attached, not just a warning letter. What happens next There's also a four-month cushion buried in the text. AI systems already on the EU market before today get until December 2, 2026 before they must meet the watermarking requirement under Article 50 2 , the rule requiring machine-readable markers embedded in AI-generated content. That's a genuine grace period, but it's a narrow one, and it doesn't touch the disclosure duties that are binding as of this morning. The timing lines up with a broader shift outside the EU too. California's AI transparency law, which requires similar disclosures for chatbots and AI-generated content, is moving on a parallel track this year. Regulators on both sides of the Atlantic are converging on the same basic demand: tell the user when they're talking to a machine. The EU just got there first, and it's the one with fines already switched on. Also read: Anthropic's COBOL Tool Rattled IBM, But AI Migration Still Hides Silent Bugs https://startupfortune.com/anthropics-cobol-tool-rattled-ibm-but-ai-migration-still-hides-silent-bugs/ • Retail Traders Are Building AI Trading Bots Once Reserved for Hedge Funds https://startupfortune.com/retail-traders-are-building-ai-trading-bots-once-reserved-for-hedge-funds/ • Alibaba Says Its New Qwen3.8-Max Model Trails Only Anthropic's Claude https://startupfortune.com/alibaba-says-its-new-qwen38-max-model-trails-only-anthropics-claude/