AI Omnibus Comes into Effect in Europe and Imposes 22-Day Deadline for Transparency in Corporate Chatbots
With the package approved by the European Council on June 29, the horizon for high-risk AI extends to December 2027, but the deadline of August 2 for transparency in chatbots remains unchanged.
The Digital Omnibus AI package, approved by the Council of the European Union on June 29, 2026, following the European Parliament's approval on June 16, came into force this week with publication in the EU Official Journal, consolidating the most significant revision to the compliance timeline of the European AI Act since its original implementation in August 2024. Companies now have definitive dates to build their compliance plans, but the margin for maneuvering is asymmetric: for high-risk AI, there are 17 additional months; for chatbots and general-purpose language models, the clock reads August 2, 2026.
What Changed and What Remained the Same
The central change of the Omnibus extends the compliance deadline for high-risk AI systems listed in Annex III of the law, including credit screening tools, employee selection, and risk assessment in insurance, from August 2026 to December 2, 2027. For high-risk systems embedded in regulated products, such as medical equipment or autonomous vehicles, the deadline moves to August 2, 2028.
What the Omnibus did not change is Article 50: the obligation for providers to inform users when they are interacting with an AI system. Chatbots, virtual assistants, and customer service agents must be compliant by August 2, 2026. Any company operating in the European space with such tools that has not implemented AI identity marking is on a 22-day countdown.
Starting August 2, the European Commission will also gain formal powers to impose retroactive fines for violations committed since August 2025, specifically for providers of general-purpose language models (GPAI). Models trained with computational capacity above 10^25 FLOPs will be subject to obligations for technical documentation, copyright policies, and systemic risk assessments. Twenty-four organizations, including Amazon, Anthropic, Google, IBM, Microsoft, and Mistral AI, have signed the Code of Good Practice for GPAI published by the European AI Office.
New Prohibitions in Article 5
The package adds two prohibitions to Article 5 of the law: AI systems used to generate child sexual abuse material or non-consensual intimate content, including tools for generating intimate images of real people without consent, are prohibited both in marketing and use. For companies developing image generation products, the ban is immediate with the entry into force of the Omnibus.
The package also sets December 2, 2026, as the deadline for systems already on the market when the original law came into force to adopt obligations for marking AI-generated content, including watermarking and automated disclosure.
Impact on American and Japanese Companies Operating in the EU
For major providers of general-purpose language models in the United States, OpenAI, Google, and Anthropic, the August 2 deadline is the most concrete milestone in the short term. They must have completed systemic risk assessments of models trained above the threshold of 10^25 FLOPs, under penalty of fines that the Commission can now impose retroactively for violations dating back to August 2025.
In Japan, NEC, Fujitsu, and Sony AI, which operate AI services in European markets, fall under the same regulatory scope. Fujitsu, with generative AI operations in several EU countries, and NEC, with digital government contracts on the continent, will need to demonstrate compliance with GPAI obligations if their models exceed the established computing threshold.
For global consulting firms, the Omnibus reorients the immediate delivery agenda. Accenture, Capgemini, Deloitte, and PwC were already marketing compliance services with the AI Act; the extension of deadlines for high-risk AI frees up resources that can be redirected to compliance for GPAI and chatbots, now the most urgent point on the European regulatory calendar. Clients with deployments of AI agents and virtual assistants in EU territories are set to become a priority for these firms' compliance teams in the next three weeks.
The August 2 deadline is non-negotiable. For companies that have not yet documented how their AI systems interact with users in Europe, the compliance window closes in 22 days without the possibility of further extension.