The European regulation on artificial intelligence, the AI Act, entered into force in August 2024, with obligations that apply in stages. For many SME leaders, it remains unclear: a long, technical text, often presented either as a threat or as something only for big tech.
The reality is in between. Most SMEs are not affected by the heaviest obligations, but almost all have something to do. This article sets out what really concerns you. It does not replace legal advice, but it tells you which questions to ask.
A risk-based approach
The AI Act does not regulate “AI” in general: it regulates uses, ranked by the risk they pose to people.
- Prohibited practices: manipulation exploiting people’s vulnerabilities, social scoring, emotion recognition at work or in schools, certain forms of biometric identification. These bans have applied since February 2025.
- High-risk systems: uses that can significantly affect people’s lives, such as recruitment and employee management, creditworthiness assessment, access to education or essential services. They carry the heaviest obligations.
- Systems subject to transparency: chatbots, content generated or altered by AI. The main obligation is to inform people. It has applied since 2 August 2026.
- Minimal-risk uses: the vast majority of business uses, such as writing assistance, document summarization, ticket classification or invoice processing. No specific obligation beyond the common baseline.
Provider or deployer: what is your role?
The regulation distinguishes two main roles. The provider develops an AI system and places it on the market. The deployer uses an AI system in the course of its professional activity.
An SME using an off-the-shelf assistant, a résumé screening tool or a vendor’s chatbot is usually a deployer. An SME that develops its own system and offers it to customers, or substantially modifies an existing system, may become a provider, with much heavier obligations. Identifying your role for each system is the first step.
The question is not “do we use AI?”, but “for what use, in which role, and at what risk level?”.
What applies to almost every company
- AI literacy: the regulation asks providers and deployers to ensure that people using AI systems have a sufficient understanding of how they work and of their limits. In practice: training suited to real use, and a clear usage policy.
- Transparency: a customer interacting with a chatbot must know they are talking to an AI, unless it is obvious. AI-generated content, especially realistic images, video or audio, must be flagged as such in the cases set out by the text.
- Respecting the bans: checking that no tool in use falls under a prohibited practice, for example a tool analyzing employees’ emotions during video calls.
If you use a high-risk system
The most common case in SMEs is recruitment: a tool that screens, ranks or scores applications falls into the high-risk category. As a deployer, you must in particular:
- use the system in line with the provider’s instructions;
- assign oversight to competent people, able to understand its outputs and challenge them;
- ensure the data you feed it is relevant;
- monitor its operation and report serious incidents;
- keep the logs generated by the system;
- inform employees and their representatives before using it in the workplace, and inform the people concerned that a decision about them relies on such a system.
The Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026) moved the timeline. Obligations for the Annex III high-risk systems, including recruitment, apply from 2 December 2027. Obligations for AI systems built into regulated products (Annex I) apply from 2 August 2028.
This delay is no reason to wait: it is the time you need to inventory your tools, obtain documentation from your providers and train the people in charge of oversight.
What about the GDPR?
The AI Act does not replace the GDPR: both apply in parallel. An AI recruitment tool must therefore meet both AI Act and GDPR obligations, in particular on legal basis, informing candidates, impact assessment and safeguards for automated decisions. In practice, a serious GDPR approach already covers a good part of the way.
The checklist to get started
- Take inventory of the AI tools used in the company, including those adopted without approval.
- Classify each use: prohibited, high-risk, transparency, minimal risk.
- Identify your role for each: deployer or provider.
- Train teams on the tools they use and their limits.
- Update the information given to customers, candidates and employees when AI is involved.
- Require documentation from providers for any high-risk system.
Our recommendation
Do not let the AI Act paralyze your projects: most useful business uses are minimal risk. But do not ignore it either. An inventory of your uses and basic team training already cover most of the obligations that apply to you. For sensitive uses, such as recruitment or customer assessment, build compliance into the design from the start, with your DPO or legal counsel: it is much simpler than adding it afterwards.