The EU AI regulation has applied generally since 2 August 2026, including its transparency rules: a chatbot that answers tenants must tell them they are talking to an AI. The omnibus regulation of 8 July 2026, however, postponed the obligations for “high-risk” systems to 2 December 2027. Here is what a property management agency must check now, and what can wait.

What applies since 2 August 2026?

Regulation (EU) 2024/1689, known as the AI Act, entered into force on 1 August 2024. Its Article 113 sets its general date of application: 2 August 2026. Some parts already applied, such as the AI literacy obligation (Article 4) since 2 February 2025. The transparency rules of Article 50 are among the provisions that become applicable on 2 August 2026.

Regulation (EU) 2026/1744 of 8 July 2026, known as the “digital omnibus on AI”, was published in the Official Journal on 24 July 2026. It entered into force on the third day after, 27 July 2026. It mainly postpones the obligations for high-risk systems. Those under Annex III will apply from 2 December 2027, those under Annex I from 2 August 2028. It does not postpone Article 50, whose paragraphs 1 to 6 remain unchanged. For an agency, the timetable therefore has two stages. Transparency towards tenants and applicants applies now. Any heavy obligations linked to high risk will wait until the end of 2027.

What role does your agency play under the regulation?

The regulation allocates obligations by role. The provider is the one that develops an AI system, or has it developed, and places it on the market or puts it into service under its own name. The deployer is the one that uses it under its own authority, outside a purely personal activity. An agency using management software with AI features is in principle a deployer. If it develops a chatbot itself, or has one developed under its name, it becomes the provider of that system. Four situations are the most common:

  • purchased software that sorts or summarises requests internally;
  • a general-purpose tool the team uses for drafting;
  • a chatbot or voice assistant exposed to tenants or applicants;
  • a tool that assesses the creditworthiness of rental applicants.

Each calls for different checks, detailed below.

Must your tenant chatbot present itself as an AI?

Yes. Article 50(1) requires AI systems intended to interact directly with natural persons to be designed so that those persons are informed that they are interacting with an AI system. Legally, this obligation rests on the provider, which must design the system accordingly. But it is the agency that puts the chatbot in front of its tenants: it must check that the tool it deploys announces itself as an AI from the first exchange. The information must be given at the latest at the time of the first interaction (paragraph 5). For deployers, paragraph 4 adds a rule on text. AI-generated text published to inform the public on matters of public interest must be disclosed as such, unless it has undergone human review and a person holds editorial responsibility for it. In practice: a chatbot visible to tenants introduces itself, and published content is reviewed and owned by a person.

What about marking generated content?

Article 50(2) requires providers of systems that generate synthetic content to mark it in a machine-readable format. The omnibus regulation gives systems placed on the market before 2 August 2026 until 2 December 2026 to comply. This is a provider obligation, not an agency one. It concerns the agency only if it developed such a tool itself.

Is your request-sorting tool “high-risk”?

In principle, no. High-risk systems under Annex III are listed by area. Point 5, on access to essential private and public services, covers in particular systems intended to evaluate the creditworthiness of natural persons or establish their credit score. It does not mention sorting maintenance requests, prioritising tickets or drafting replies. The word “housing” appears in no article or annex of the regulation: it is cited only in the recitals. Recital 58 explains that credit scores determine access to essential services such as housing. For an agency, the line is therefore drawn by use. A tool that classifies water leaks and suggests a plumber does not fall under Annex III. A tool that assigns a creditworthiness score to rental applicants could. In that second case, have the use reviewed by a lawyer before December 2027.

Applicant selection remains governed by regional law and anti-discrimination rules, which already apply. Our article on income of three times the rent sets out the accepted criteria, and the one on false pay slips the useful checks.

What must you show about staff training?

Article 4 has imposed an AI literacy obligation since 2 February 2025. Its original version required providers and deployers to take measures to ensure, to their best extent, a sufficient level of AI literacy. The omnibus regulation rewrote it. They now take measures to support the development of AI literacy among their staff. The text specifies that this obligation does not require them to guarantee any specific level of AI literacy for an individual. The obligation remains, but it becomes an obligation of means. For an agency, the evidence of these measures consists of a few simple documents. A list of the tools used and their uses. A written instruction on what may be entered in them. A record of the team’s onboarding sessions. No certificate is required, and no “AI Act certification” exists for software.

Who supervises the regulation in Belgium?

According to an answer to a written question in the Senate of 16 January 2025, the FPS Economy currently coordinates the implementation of the AI Act in Belgium. At the time of writing, we have found no text published in the Belgian Official Gazette designating the market surveillance authority. On 15 December 2025, the FPS Economy launched an information campaign with a self-assessment tool, aiactpourmoi.be. The release noted that many companies, especially SMEs and start-ups, are not yet well informed. The tool helps a company situate itself, but it is neither legal advice nor a compliance decision.

What checklist for your agency?

  1. List the AI tools used in the agency, who uses them and for what.
  2. Classify each use: internal (sorting, summarising, drafting), exposed to tenants or applicants (chatbot, voice assistant), or linked to creditworthiness.
  3. Check exposed chatbots: they announce themselves as an AI from the first exchange.
  4. Review and own any AI-generated text before publication.
  5. Isolate creditworthiness scoring tools and have them examined by a lawyer before December 2027.
  6. Document staff training and the written usage instruction.

On keeping evidence of exchanges and findings, see also our article on the digital inventory.

Where does Seido’s AI triage stand?

Managing several properties? Seido uses AI to sort incoming requests internally: it suggests a category, a priority and a contractor. Depending on the level of autonomy you set, it waits for your approval on what commits the agency: the AI suggests, you decide. This is an internal sorting use, to be recorded in your inventory with its purpose, the data processed and the planned human oversight. See how Seido’s AI triage works.

Frequently asked questions

Have AI Act penalties applied since 2 August 2026?

The regulation’s chapter on penalties has applied since 2 August 2025, according to Article 113. It requires Member States to lay down the rules on penalties and other enforcement measures (Article 99). Since 2 August 2026, the transparency obligations of Article 50 are among the applicable rules, and can therefore be sanctioned. For high risk, the obligations themselves apply only from 2 December 2027 or 2 August 2028. In Belgium, actual supervision also depends on the designation of the authorities, which we did not find published at the date of this article.

Must an e-mail drafted with ChatGPT mention AI?

Not under Article 50 in most cases. The obligation to disclose generated text concerns text published to inform the public on matters of public interest, and it falls away if a person has reviewed the text and holds responsibility for it. An individual e-mail to a tenant does not fall into that case. Other rules still apply: data protection, and for real estate agents their professional obligations.

Must our agency appoint an AI officer?

The regulation does not require a deployer that uses no high-risk system to appoint a person. It requires AI literacy measures suited to the staff concerned. In a small agency, entrusting the tool inventory and usage instruction to an identified person remains the simplest way to be able to show those measures.

Is the aiactpourmoi.be tool enough to be compliant?

No. It is an information tool launched by the FPS Economy in December 2025, during a campaign aimed mainly at SMEs, to help companies find out whether they are concerned. It gives a first diagnosis, not a compliance decision. For a sensitive use, such as assessing applicants’ creditworthiness, a lawyer’s advice remains necessary.

Sources and references

Last updated: August 2026. This article is provided for information purposes only and does not constitute legal advice. For any specific situation, consult a specialised lawyer.