Since 2 August 2026 the EU AI Act has required chatbots to say they are machines and deepfakes to be labelled.
Since 2 August 2026 the transparency rules of the EU AI Act have applied across the European Union. A chatbot has to tell the person on the other side that they are talking to a machine, and realistic images and videos created or altered with AI must be labelled. This reaches anyone who uses such technology in their work with property buyers and sellers. According to the European Commission, the requirements for high-risk systems, such as those used for credit scoring, will apply later, from 2 December 2027.
The Act entered into force on 1 August 2024, but according to the Commission its requirements take effect in stages. The bans on certain practices have applied since 2 February 2025 and the rules for general-purpose models since 2 August 2025, while from 2 August 2026 the Act applies as a whole, with some exceptions. The AI Omnibus, a package of changes to simplify it, entered into force on 27 July 2026. Under it, the rules for high-risk uses in sensitive areas (Annex III to the Act) will apply from 2 December 2027, and those for systems built into products such as lifts from 2 August 2028.
The first requirement concerns systems that talk directly to people. The provider, which developed the system and offers it under its own name, is obliged to design it so that people know they are dealing with AI, unless that is obvious. According to the European Commission's questions and answers on Article 50, people must be told this from the start of the first interaction, and the exception for obvious cases is to be interpreted restrictively. If you write to an assistant on a property website or in your bank's app, it has to tell you from the start that it is a machine.
The second requirement covers deepfakes. These are images, audio or video created or altered with AI that resemble existing persons, objects, places or events and would falsely appear to be authentic. Here the duty falls on whoever uses the system in their work. They must label the content the first time a person sees it, with a label that can be seen or heard, and cannot rely only on the provider's machine-readable marking, the Commission says. If you are buying and have doubts about the photos in a listing, ask whether they were processed with AI and whether that is labelled.
The third requirement applies to text created or altered with AI and published to inform the public on matters of public interest. Such text must be labelled unless a person has reviewed it or it has gone through editorial control. According to the Commission, checking spelling or grammar is not enough. Content created before 2 August 2026 does not need to be labelled retroactively, and providers of systems placed on the market before that date must apply machine-readable marking from 2 December 2026. Breaches of Article 50 are punishable by fines of up to 15 million euros or up to 3% of total worldwide turnover for the preceding financial year, and enforcement lies mainly with national market surveillance authorities.
The European Commission lists credit scoring that can deny a person a loan among the high-risk uses of AI. The rules for these uses will apply from 2 December 2027. From that date, such systems will have to meet strict requirements before they reach the market: risk assessment and mitigation, high-quality data to reduce the danger of discriminatory outcomes, logging of their activity, detailed documentation, clear information to the institution that uses them, appropriate human oversight and a high level of robustness, cybersecurity and accuracy.
Under Article 26 of the Act, a bank that uses such a system to make a decision about an applicant, or to help it make one, must tell the applicant that a high-risk AI system is being used in their case. Human oversight must be assigned to people with the necessary competence, training and authority. These duties apply from 2 December 2027, together with the other rules for high-risk uses. If you apply for a mortgage before then, you can ask the bank whether an AI model plays a part in assessing your creditworthiness and who reviews its result.
According to BTA, the national coordinator for applying the Act in Bulgaria is the Communications Regulation Commission (КРС). On 2 September the Minister of Innovation and Digital Transformation, Ivan Vasilev, told the relevant parliamentary committee that work on Bulgaria's AI law was nearing its end and that the next step was a public consultation on the text, BTA reported.
Meanwhile, on 24 September ten MPs filed a bill on the use and development of artificial intelligence with the National Assembly. Parliament's record shows that the lead committee on it is the Committee on Innovation and Digital Transformation, and that ministries, state agencies and institutes have sent opinions for its first reading, the latest on 29 October.
There is no chatbot or AI adviser on the MM site. If you are buying, the MM Tools page has four tools for you, and two of them, MM Property Check and Contract Read, use AI. In the first you can paste the link to a listing from one of the main listings sites, and an AI model reads its text and writes a short summary, observations and questions for the seller. It does not put a figure on what the home is worth and does not read the photos. The second checks a preliminary contract, a notarial deed of sale or a Property Register document against a checklist and marks which points it found, which are unclear and which it did not find. It is not legal advice: the tool can miss something or misread the text, and the uploaded file is deleted as soon as the reading starts. The Real Cost and the Mortgage Calculator do not use AI: the first estimates the main costs of buying a property in Sofia from the published tariffs and rules, and the second applies the Bulgarian National Bank's lending rules and approves no loan — the bank decides.
If you are selling, AI Price Range gives an indicative price range, not a valuation. It shows a from–to range in euros from the asking prices that Bulgarian property portals publish, gives the date of its figures and never gives one price. AI writes only the short explanation, and the range itself is worked out in code. Sale Dossier draws up, from a table of rules, the list of documents a seller brings to the notary. AI takes part only if you upload your title deed, and then it fills in two of the answers, which you confirm. An MM consultant also prepares a free written valuation, with the comparable sales it rests on, when you ask for one on the Sell page.
If you are building or investing in construction, MM has no tool made for you, with AI or without it. Here MM uses AI only for translation: descriptions of new-build projects that the office writes in Bulgarian are translated into English, and no person reviews each translation. The new-build page works without AI. For a project MM sells, it shows the building floor by floor with its available apartments and the stage construction has reached. A developer can contact MM through the partnership option of the contact form on the About page.
From 2 August 2026. Providers of systems placed on the market before that date must apply machine-readable marking to the content they generate from 2 December 2026.
Yes. The provider must design it so that people learn from the start of the first interaction that they are dealing with AI, unless that is obvious. The European Commission says that exception is to be interpreted restrictively.
The European Commission lists credit scoring among the high-risk uses of AI. After the AI Omnibus, the rules for these uses will apply from 2 December 2027.
On 2 September 2026 the Minister of Innovation and Digital Transformation said work on the law was nearing its end and that a public consultation would follow. A separate bill by ten MPs, filed on 24 September, has received opinions for its first reading in the National Assembly.