AI Act Delayed? What Still Changes for Your SME on 2 December
In July, the European Union adopted the Digital Omnibus, which postponed part of the EU Artificial Intelligence Act (AI Act). Many headlines summed it up in one word: delayed. For most SMEs, that reading is risky.
In this article9 sections
What was delayed are the obligations for high-risk systems. The transparency rules, which affect any company with a chatbot or that publishes AI-generated content, have applied since 2 August 2026. And on 2 December 2026 new deadlines arrive. This article covers what changed, what did not, and what to do in the coming weeks.
What the Digital Omnibus changed
The Digital Omnibus is Regulation (EU) 2026/1744, which amends the AI Act. According to Orrick's analysis, the main timeline changes are:
- Annex III high-risk (recruitment and people management, credit scoring, education, access to essential services): obligations move to 2 December 2027.
- Annex I high-risk (AI built into already regulated products, such as machinery or medical devices): moves to 2 August 2028.
- Smaller companies: the simplifications for SMEs, such as simplified documentation and priority access to regulatory sandboxes, are extended to a new category of mid-sized companies.
The nature of the change matters: it is a postponement, not a repeal. The obligations remain, with more time to meet them.
What was not delayed
- Transparency (Article 50), since 2 August 2026. Anyone talking to a chatbot or voice assistant must know they are dealing with an AI, unless it is obvious. Anyone publishing deepfakes (manipulated images, audio or video that look real) must disclose that the content was generated or altered by AI. Gibson Dunn confirms these rules keep their original timeline.
- Marking of generated content, by 2 December 2026. Generative AI providers must mark what their tools produce in a machine-detectable way. Systems already on the market before August have until 2 December 2026.
- The original prohibited practices, since February 2025. For example, manipulation that exploits vulnerabilities, social scoring and emotion recognition in the workplace.
Technical marking is mainly the tool provider's obligation. But your company chooses the providers, and it is your company that is exposed if it publishes content that misleads customers.
Is your SME affected? Five questions
- 1Do you have an AI chatbot or assistant on your website, on WhatsApp or on the phone? The transparency rules already apply: customers must know they are talking to an AI.
- 2Do you publish AI-generated images, video, audio or text? Check whether any of it could pass for real and whether your tool marks the content.
- 3Do you use AI to screen candidates, assess credit or make decisions about people? That is high risk. You have until December 2027, but preparation takes months.
- 4Do you build or sell a product with AI? Then you are a provider and the obligations are greater. A specific review is worthwhile.
- 5Does your team use AI tools without rules? The AI Act is not the only risk: there is also the GDPR and confidentiality. See Shadow AI: the invisible risk.
If you answered "no" to all of them, the direct impact is small. If you answered "yes" to any, follow the checklist.
Checklist before 2 December
- 1Take inventory. List the AI tools used in the company, what for and with which data. Include the ones your team uses on its own.
- 2Review customer touchpoints. In every chatbot, voice assistant or automated reply, make sure there is a clear sentence such as "You are talking to a virtual assistant".
- 3Review published content. AI-generated images and videos that could look real should be labelled as such.
- 4Ask your providers. How do they handle content marking and transparency? Get the answer in writing.
- 5Write a short policy. One page is enough: approved tools, data that must not be shared, who reviews what gets published.
- 6Train your team. The AI Act still requires companies to take measures to support the AI literacy of the people who use these systems, although it no longer sets a minimum level. A team that understands AI's limits makes fewer mistakes.
What to prepare by December 2027
If you use AI in recruitment, performance reviews, credit or access to services, use the extra time:
- Human oversight: decide who reviews decisions and when they can override the tool.
- Records: keep what the tool decided and with which data.
- Provider: ask for the technical documentation and the declaration of conformity once available.
- Informing people: candidates and customers should know when a decision is supported by AI. The GDPR already requires this in many cases; see GDPR and Artificial Intelligence.
What is at stake
The AI Act provides for fines of up to €35 million or 7% of worldwide turnover for prohibited practices, and up to €15 million or 3% for other obligations, such as transparency. For SMEs, the lower of the two applies. These are ceilings, not typical amounts, but they show the rules are meant to be taken seriously.
For an SME, the most immediate risk is a different one: trust. A customer who finds out they were talking to a machine without knowing it, or who sees a fake image linked to your brand, is unlikely to come back.
How TecLab can help
With our AI Governance and Compliance service, we inventory your tools, check transparency at customer touchpoints and help you write your usage policy. We start with whatever has the nearest deadline.
Conclusion: delayed does not mean on hold
The Digital Omnibus gave more time to companies that use AI in decisions about people. For everyone else, the rules that matter most are already in force or arrive on 2 December. Start with the inventory: within a few hours you will know whether your company has anything to do.
This article is for information only and is not legal advice.
Frequently asked questions
Written by
Teclab Team
Artificial intelligence and automation consultants who help SMEs work better.