- Lead. On August 2, 2026 — 33 days from today — the European Commission gains the power to impose financial penalties on general-purpose AI model providers for the first time under the EU AI Act, activating fines of up to €15 million or 3% of global annual turnover for non-compliance.
- Fact. The same date triggers Article 50 transparency obligations requiring chatbots and AI assistants to disclose at first interaction that they are AI systems, and activates national market surveillance authorities’ full investigatory powers across all AI Act obligations.
- Stake. The Digital Omnibus agreement reached by EU lawmakers on May 7, 2026 simultaneously pushed the most burdensome obligations — compliance deadlines for high-risk AI systems — back by 16 months to December 2, 2027, in the first significant revision to the law since its adoption.
What changes on August 2
The EU AI Act has been in force since August 2024, but its enforcement architecture has been rolling out in stages. The most consequential date yet arrives on August 2, 2026. From that date, the European Commission’s AI Office gains the power to request documentation from GPAI providers, conduct model evaluations, order compliance measures, and impose fines. The maximum penalty for GPAI-related violations is €15 million or 3% of global annual turnover, whichever is higher.
Article 50 transparency obligations also activate on August 2. Under these rules, systems that interact directly with users — chatbots, AI assistants, voice agents — must notify those users at the start of each interaction that they are communicating with an AI, unless the context makes it evident. Systems that detect emotions or categorise users biometrically must also notify users during operation. A separate synthetic-content watermarking obligation — requiring machine-readable markers on AI-generated audio, images, and video in line with the C2PA standard — takes effect the same day for new deployments, with a legacy grace period to December 2, 2026 for systems already on the market.
The Digital Omnibus changes
EU legislators agreed on May 7, 2026 to the first substantial amendments to the AI Act, under the banner of the Digital Omnibus on AI. The most significant change delays the compliance deadline for stand-alone high-risk AI systems listed under Annex III — a category covering recruitment tools, credit-scoring systems, law enforcement applications, border control tools, and education technology — from August 2, 2026 to December 2, 2027. That is a 16-month extension. AI embedded in regulated products under Annex I, including medical devices and vehicles, gets a 12-month extension to August 2, 2028. The rationale offered by lawmakers was that the harmonised technical standards from European standards bodies CEN and CENELEC will not be available until Q4 2026 at the earliest, making it impossible for companies to achieve compliant conformity assessments on the original timeline.
The Omnibus also introduced two new prohibited practices, due to take effect December 2, 2026: AI systems that generate or manipulate non-consensual intimate imagery, and systems that generate child sexual abuse material. Both carry the Act’s maximum penalty tier of €35 million or 7% of global annual turnover. As Congress separately drafts US federal AI legislation targeting frontier developers, the EU’s penalty activation puts Brussels ahead of Washington in moving from policy to enforcement.
Who is covered and who has signed the code of practice
GPAI obligations — documentation, copyright compliance, training data transparency, downstream provider disclosures, and adversarial testing for the most powerful models — have technically been in force since August 2, 2025. The change on August 2, 2026 is that the Commission can now penalise failures to comply with those obligations rather than simply requesting corrective action. The AI Office’s GPAI Code of Practice, published in July 2025, counts approximately 24 signatories including Amazon, Anthropic, Google, IBM, Microsoft, Mistral AI, Aleph Alpha, Cohere, and Samsung Electronics. Meta has not signed the full code. xAI signed only the Safety and Security chapter.
The August 2 activation marks the point at which the EU AI Act transitions from a compliance framework with soft teeth into one with hard financial consequences, according to analysis by Axis Intelligence. Non-signatories of the GPAI code of practice lose the evidentiary benefit that signing confers, and face a higher evidential burden if the Commission initiates an investigation. The formal adoption and publication of the Digital Omnibus in the EU Official Journal is expected before the August 2 date.