In Europe, AI now has to introduce itself
According to the European Commission's official page, the AI Act became applicable on 2 August 2026, with some exceptions. The Article 50 transparency obligations came into play on the same date: systems that interact with people must make that clear, generated content must be marked in machine readable form, and deepfakes must be openly disclosed.
A two year transition period ran out
The European Commission's regulatory framework page on AI gives the dates plainly: the AI Act entered into force on 1 August 2024 and became applicable on 2 August 2026, with some exceptions. Referring to a press release dated 31 July 2026, the same page states that the Commission started enforcing AI Act rules and new transparency requirements on 2 August.
So this is not a new law but the arrival of the application date for a regulation adopted two years ago. The page notes that the transparency rules come into effect in August 2026 and that the framework requires disclosure when humans interact with an AI system such as a chatbot.
PPC Land covered it on August 1, reporting that Article 50 forms part of Regulation (EU) 2024/1689 and requires providers and deployers of interactive AI systems to disclose their artificial nature, mark synthetic content in machine readable formats, and notify users of emotion recognition or biometric categorisation use.
Four obligations, three dates and a penalty ceiling
The transparency obligations address four groups: interactive AI systems; providers of systems that generate or manipulate synthetic content; deployers of emotion recognition and biometric categorisation systems; and deployers publishing deepfakes and AI generated text on matters of public interest.
On the chatbot side, the rule works like this: systems that interact directly with natural persons, meaning chatbots, AI agents and avatars, must be designed and developed so that people are informed they are interacting with AI.
On the content side the obligation is technical. Outputs of generative systems must be marked with effective, reliable, robust and interoperable machine readable marks that allow them to be detected as generated or manipulated by AI. On deepfakes, Article 50(4) requires deployers publishing AI generated or manipulated image, audio or video content to clearly disclose that it is artificially generated or manipulated. Content that looks or sounds like a real person must be labelled even where no deception was intended and no real individual is depicted.
The dates come in three stages. 2 August 2026 is the start of application. 2 December 2026 ends the transitional period for providers of systems placed on the market before 2 August 2026 to meet the marking and detection obligations. PPC Land also reports 2 February 2027 as a watermark detection interoperability requirement.
On penalties, the figure PPC Land gives is up to €15 million or 3% of worldwide annual turnover, whichever is higher, with lower thresholds for SMEs. The Commission page we reviewed states no penalty amounts, so that figure belongs to PPC Land.
Disclosure is the easy half of the job
The most visible face of the rule is adding one sentence: you are talking to an AI. That is technically simple. The hard part sits on the marking obligation: making generated content machine readably marked is a requirement about the production pipeline itself. Whether the system that produced an image or a piece of text carries that mark is not something an agency can bolt on afterwards.
The assessment PPC Land relays points the same way. In the words of Cinchy CEO J. Paul Haynes, the disclosure requirement is the easier half of a harder problem; the real issue is continuous operational oversight of AI systems.
The second key point is breadth of scope. The rule binds not only those building models but the parties deploying them. A brand publishing a deepfake, a retailer using emotion recognition or a business running a bot that talks to customers is under obligation even without being a developer.
Third, this regulation sets a direction. Labelling AI generated content is no longer one market's preference; the commercial advertising regulation that took effect in Türkiye on 1 August 2026 also includes AI disclosure. The two texts are not the same, but they face the same way. That comparison is our own comment.
What it means for businesses in Türkiye
The assessment below is our reading, not something stated in the sources. None of them contains a provision or breakdown relating to Türkiye; the regulation belongs to the European Union.
Even so it bears directly on four situations. First, selling into the EU: if you run a site, app or chatbot serving users in Europe, scope follows where the user is. Second, exports and marketplaces: if a brand selling on EU marketplaces produces product imagery with AI, the marking obligation becomes part of your supply chain. Third, the client chain: if you provide agency services to a European client, the content you produce will be expected to meet these requirements. Fourth, the local parallel: the commercial advertising rules in Türkiye also call for AI disclosure, so meeting both through a single content discipline is cheaper than treating them separately.
The practical first step is simple: record which content was produced with AI. A team keeping no production log will not know six months later which image needs labelling. That last sentence is our own comment.
The UNALSOFT take
In our AI ad film production, how each shot was produced is logged from the start. The reason is not an expectation of regulation but the nature of the work: if nobody knows where a delivered creative came from, it can neither be fixed nor defended. Article 50 turns that habit into a requirement. The question to ask is not whether to label, but whether the production log is orderly enough to generate the label correctly.
Does your content have a paper trail?
Let's set up the record and the disclosure for AI generated content together.