Next Friday I start reading the AI Act one article at a time, and a single article is hard to follow without knowing where it sits, so this week is an overview of both regimes. The EU has one regulation of 113 articles that applies in every member state. The UK regulates AI through the laws and regulators it already had. Software sold in both has to satisfy both.
The EU’s regulation is Regulation (EU) 2024/1689, amended in July by the Digital Omnibus, Regulation (EU) 2026/1744. What it asks of you depends on who you are, what your system does, and when.
Who you are
Almost every duty in the Act is addressed to a provider or a deployer, so the two definitions in Article 3 decide most of what follows. Here they are, verbatim, in the format every article gets from next week. Select any part of the text and its plain reading and notes open underneath.
EU AI Act · Article 3, Definitions
‘provider’ means a natural or legal person, public authority, agency or other body that develops an AI system or a general-purpose AI model or that has an AI system or a general-purpose AI model developed and places it on the market or puts the AI system into service under its own name or trademark, whether for payment or free of charge;
‘deployer’ means a natural or legal person, public authority, agency or other body using an AI system under its authority except where the AI system is used in the course of a personal non-professional activity;
Select any part of the text. 8 clauses, 8 notes.
Article 3(3)
Anyone can be a provider
‘provider’ means a natural or legal person, public authority, agency or other body
What it means
A company, a public body, an agency or a single person. Size and sector do not come into it, and neither does where the provider is established: Article 2(1)(a) applies the Act to providers “irrespective of whether those providers are established or located within the Union or in a third country”.Dates
In force since 2 February 2025, and untouched by the Omnibus
Article 3(3)
Builds it, or has it built
that develops an AI system or a general-purpose AI model or that has an AI system or a general-purpose AI model developed
What it means
The provider is whoever develops the AI system or the general-purpose model, or commissions someone else to develop it. Paying an agency to build your chatbot leaves you as the provider. The definition covers two different things, a system and a model, and they carry different duties.Software
Calling someone else's model does not stop you being a provider
The definition asks who developed the AI system. It does not ask who trained the model inside it. A product that sends prompts to a third-party model API and shows people the result is an AI system, and the team that built the product developed it.
The Commission’s guidelines on Article 50 give exactly this as their first example of a provider: “a company [that] provides a generative or interactive AI application (e.g. a chatbot, image generator, AI agent) on the Union market under its own name or trademark”. The model vendor is the provider of a general-purpose AI model, with separate duties under Articles 53 to 55.
Article 3(3)
Puts it on the EU market or into service
and places it on the market or puts the AI system into service
What it means
Developing something is not enough. The provider is the one who first makes it available on the Union market (placing on the market) or supplies it for first use, including its own use (putting into service). A model can only be placed on the market. A system can be either.Elsewhere in the law
The two verbs are defined a few points further down
Placing on the market is “the first making available of an AI system or a general-purpose AI model on the Union market” (Article 3(9)). Putting into service is “the supply of an AI system for first use directly to the deployer or for own use in the Union for its intended purpose” (Article 3(11)).
“For own use” is the part in-house teams miss. An organisation that builds a system and uses it itself has put it into service, and is both its provider and its deployer.
Elsewhere in the law
A provider outside the EU is still caught if the output is used there
Article 3(3)
Under its own name or trademark
under its own name or trademark
What it means
The name on the product decides who the provider is. Whoever presents the system to the market as theirs carries the provider’s duties, whoever wrote the code or trained the model.Elsewhere in the law
Putting your name on a high-risk system makes you its provider
Article 3(3)
Paid or free
whether for payment or free of charge
What it means
Giving it away does not take you out of the definition. A free tier, a public demo and an internal tool are all capable of making you a provider.Elsewhere in the law
Free of charge is in scope, but open source has its own exclusion
Price is irrelevant to the definition. Licence is handled separately: under Article 2(12) the Act “does not apply to AI systems released under free and open-source licences, unless they are placed on the market or put into service as high-risk AI systems or as an AI system that falls under Article 5 or 50”.
So an open-source system that chats with people or generates content is still inside Article 50, and an open-source system used for one of the high-risk purposes is still high-risk.
Article 3(4)
Anyone can be a deployer
‘deployer’ means a natural or legal person, public authority, agency or other body
What it means
The same list of bodies as for a provider. A deployer is defined by what it does with a system, so one organisation can be the provider of the system it sells and the deployer of the systems it buys.Article 3(4)
Uses an AI system under its authority
using an AI system under its authority
What it means
The deployer is the organisation that decides to use the system and is responsible for that use. An employee operating the tool at work is acting under the employer’s authority, so the employer is the deployer. The definition speaks of AI systems only. There is no such thing as the deployer of a model.Software
What a deployer has to do, and from when
A deployer’s duties depend on what the system is. For any AI system, Article 4 asks the deployer to take measures to support the AI literacy of its staff. For emotion recognition, biometric categorisation, deep fakes and AI-written public-interest text, Article 50(3) and (4) put the telling and the labelling on the deployer, and have applied since 2 August 2026. For a high-risk system, Article 26 adds use according to the instructions, named human oversight, monitoring and keeping the logs, from 2 December 2027 for the Annex III uses.
In a multi-tenant product the practical consequence is a field: which customer organisation is the deployer for this request, and which of these duties has it been told about.
Article 3(4)
Private use at home is out
except where the AI system is used in the course of a personal non-professional activity
What it means
Someone using an AI system for a personal, non-professional activity is not a deployer and carries none of the deployer’s duties. The moment the same person uses the same tool for work, the exception no longer applies.Elsewhere in the law
Article 2(10) says the same thing from the other side
Text from Regulation (EU) 2024/1689, the AI Act, consolidated with the Omnibus, checked 12 November 2026. Engineer's reading, not yet reviewed by a lawyer.
You can be a provider without training a model. If your product calls a model API and shows the result to people under your name, you developed an AI system and put it on the market. The company that trained the model is the provider of a general-purpose AI model, which is a separate role with separate duties.
There are four more roles. An authorised representative is a person or company in the EU with a written mandate to act for a provider outside it. An importer brings a non-EU company’s system onto the EU market. A distributor is anyone else in the supply chain who makes a system available. A product manufacturer puts an AI system on the market inside its own product. The Act calls all six “operators” (Article 3(8)).
What the system does
The Act sorts AI systems by what they are used for, and the duties rise with the risk.
- Prohibited practices (Article 5). Banned outright since 2 February 2025: manipulative techniques that cause harm, social scoring, untargeted scraping of faces, emotion recognition at work and in schools, and several more. The Omnibus adds two more: systems that generate non-consensual intimate imagery, and systems that generate child sexual abuse material.
- High-risk systems (Article 6). There are two ways a system becomes high-risk. The first is being a safety component of a product already regulated under the laws in Annex I, such as medical devices, machinery and toys. The second is a use listed in Annex III, which has eight areas, among them biometrics, employment, education, law enforcement and access to essential services, which is where credit scoring sits. Most of the Act’s length is about these systems: risk management, data governance, logging, human oversight, technical documentation and a conformity assessment before the system goes on the market.
- Transparency duties (Article 50). Chatbots have to say they are AI, generated content has to carry a machine-readable mark, and deep fakes have to be labelled. This applies whatever the risk tier.
- General-purpose AI models (Articles 51 to 55). Duties on the companies that train the models: technical documentation, a summary of the training content, a copyright policy, and more for models classified as carrying systemic risk.
- Everything else. No obligations beyond AI literacy (Article 4), with voluntary codes of conduct encouraged under Article 95.
The fines follow the same order (Article 99). A prohibited practice costs up to €35 million or 7% of worldwide annual turnover, whichever is higher. Most other breaches, Article 50 included, cost up to €15 million or 3%. Giving a regulator incorrect or misleading information costs up to €7.5 million or 1%. For SMEs, and since the Omnibus for small mid-caps too, it is whichever figure is lower.
When it applies
The Act arrives in stages, and the Omnibus moved the biggest one. The high-risk obligations for Annex III systems were due on 2 August 2026 and now apply from 2 December 2027.
| Date | What applies |
|---|---|
| 2 February 2025 | The prohibitions in Article 5, and AI literacy in Article 4 |
| 2 August 2025 | Duties for general-purpose AI models, the governance bodies, and penalties |
| 27 July 2026 | The Omnibus in force. Article 4 now asks providers and deployers to “take measures to support the development of AI literacy” and no longer to ensure it |
| 2 August 2026 | The Act’s general application date, which brings in Article 50, and the Commission’s power to fine model providers |
| 2 December 2026 | The end of the grace period for marking content from generative systems already on the market on 2 August, and the two new prohibitions |
| 2 August 2027 | Each member state has at least one AI regulatory sandbox running. General-purpose models already on the market before August 2025 have to comply |
| 2 December 2027 | High-risk obligations for Annex III systems |
| 2 August 2028 | High-risk obligations for systems in Annex I products |
| 2 August 2030 | Older high-risk systems intended for use by public authorities have to comply |
The Commission fills in the detail with guidelines and codes of practice. For Article 50 there are guidelines published on 20 July 2026 and a Code of Practice finalised on 10 June. For general-purpose models there is a code of practice from July 2025. The guidelines on what counts as high-risk under Article 6 were published as a draft on 19 May 2026 and were still a draft when I checked on 12 November.
Who enforces it
Each member state names market surveillance authorities, and they enforce the Act for most systems. The AI Office, inside the Commission, has exclusive power over general-purpose AI models. The Omnibus gave it more: it is now also the only enforcer for AI systems built on a general-purpose model by the same provider, with exceptions that include products under Annex I and critical infrastructure, and for AI systems that are part of a very large online platform or search engine (Article 75(1)). A company that builds on its own model and sells across the EU is regulated by the AI Office rather than a national authority.
The UK has no AI act
The UK’s approach comes from a white paper of March 2023, A pro-innovation approach to AI regulation. It gives existing regulators five principles to apply in their own sectors: safety, security and robustness; appropriate transparency and explainability; fairness; accountability and governance; contestability and redress. The principles are not law. The ICO applies them to personal data, the FCA to financial services, the CMA to competition, Ofcom to communications and the MHRA to medical devices, each using the powers it already had.
A bill has been discussed for two years and has not appeared. The King’s Speech of 13 May 2026 contained no AI bill. Parliament’s Joint Committee on Human Rights recommended one in September, and this month the AI minister said the government would look at regulation for the most advanced models if it turned out to be needed. I could find no bill, consultation or timetable.
Three things in the UK are more definite: the new rules on automated decisions, the AI Security Institute and copyright.
Automated decisions. Section 80 of the Data (Use and Access) Act 2025 replaced Article 22 of the UK GDPR with Articles 22A to 22D, in force since 5 February 2026. They cover decisions “based solely on automated processing”, meaning with “no meaningful human involvement”, that have a legal or similarly significant effect on the person. Article 22B restricts such decisions when they use special category data. Article 22C sets the safeguards for the rest: tell the person, let them make representations, let them obtain human intervention, and let them contest the decision. The ICO’s updated guidance went out for consultation on 31 March 2026 and has not been published in final form. For an engineer, 22C is the closest thing the UK has to an AI Act article.
The AI Security Institute. It tests frontier models before and after release and publishes research. It was renamed from the AI Safety Institute in February 2025 and is now part of the Cabinet Office. It is not set up in law and has no powers over anyone.
Copyright. UK law has no exception for commercial text and data mining, so training on copyrighted work for commercial use needs a licence. The Data (Use and Access) Act required the government to report on copyright and AI, and the report and impact assessment came out on 18 March 2026. They changed nothing in the law, and the government no longer has a preferred option.
Where the two meet
The AI Act applies to UK companies with EU customers. Article 2(1)(a) covers providers who place a system on the EU market “irrespective of whether those providers are established or located within the Union or in a third country”, and Article 2(1)(c) covers providers and deployers in a third country “where the output produced by the AI system is used in the Union”. A product built in the UK with customers in Germany is covered.
The part of the Act most of those products meet first is Article 50, which has applied since August. This is its first paragraph, annotated the same way as the definitions above. The whole article follows next Friday.
EU AI Act · Article 50, Transparency obligations for providers and deployers of certain AI systems
Providers shall ensure that AI systems intended to interact directly with natural persons are designed and developed in such a way that the natural persons concerned are informed that they are interacting with an AI system, unless this is obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect, taking into account the circumstances and the context of use. This obligation shall not apply to AI systems authorised by law to detect, prevent, investigate or prosecute criminal offences, subject to appropriate safeguards for the rights and freedoms of third parties, unless those systems are available for the public to report a criminal offence.
Select any part of the text. 5 clauses, 7 notes.
Article 50(1)
Providers carry the duty
Providers
What it means
The duty sits with the provider: whoever develops the system, or has it developed, and places it on the market or puts it into service under its own name or trademark, paid or free (Article 3(3)). The organisation using the chatbot is the deployer and carries no duty under this paragraph.Contested
White-labelled chatbots have no clean answer
A SaaS vendor ships a support widget, and a customer puts it on their site under their own brand. Article 3(3) hangs provider status on placing on the market “under its own name or trademark”, which points at the customer. The guidelines say an organisation becomes a provider when it modifies an existing system and puts it into service under its own name, but they do not deal with a pure rebrand of an unmodified system. Article 25(1)(a), which handles rebranding, is written for high-risk systems only.
The practical answer is to make the disclosure part of the widget and not a setting the customer can switch off. Then it does not matter which of you a regulator decides is the provider.
Article 50(1)
Systems that interact directly with people
AI systems intended to interact directly with natural persons
What it means
Four things must all be true: it is an AI system, it is intended to interact, the interaction is direct (real time or near it, including sending messages), and the other party is a natural person. Rule-based autoresponders, passive systems like facial-recognition gates, and machine-to-machine calls between agents whose outputs never reach a person are out.Software
Coding agents are in, autocomplete is out
The guidelines list “coding agents and other AI agents capable of directly interacting with natural persons” as in scope, and “AI-enabled text and code auto-completion techniques” as out. An agent that books, negotiates or answers email on someone’s behalf is in scope too, and must disclose both that it is an AI and on whose behalf it is acting.
Where a provider cannot know in advance whether its agent will meet a person, the guidelines want disclosure designed in “at the architecture level”, wherever an interaction with a person is reasonably likely, and repeated to the person instructing the agent at key steps such as authorisation and reporting.
Software
When AI drafts and a human sends, it is out of scope
Customer service staff using an AI tool to draft replies they send themselves are “indirect or mediated” interaction and fall outside paragraph 1. A bot that answers the customer directly falls inside, and “the mere possibility for humans to intervene or review” does not take it out. Where a product blends the two, the AI-written messages need disclosure unless a human properly reviewed and sent them as the main interlocutor.
In code, that means the disclosure decision belongs on each outbound message, keyed on who actually sent it, and not on the channel.
Article 50(1)
Designed in from the start
are designed and developed in such a way that the natural persons concerned are informed that they are interacting with an AI system
What it means
The person has to be informed, and the system has to be designed to do it. This is a product requirement at build time, met at the latest when the system is placed on the market or put into service, and it holds for every person who interacts with it throughout its life.Guidance
A line in the terms and conditions does not count
On their own, these are insufficient according to the guidelines: disclosure only in terms, URLs or documentation; machine-readable marks the user cannot perceive; a generic “assistant”; a site-wide “services on this website use AI”; and “this system uses LLMs” without saying what it means for the user.
What works: a first-turn statement in the conversation, a persistent label or icon, a spoken statement for voice. One prominent notice before the first interaction is “likely to suffice in most instances”, with periodic reminders in riskier settings such as financial, legal or health advice, companions, and children.
Software
The system must answer honestly when asked if it is human
The guidelines require systems to be designed to disclose “in all situations where the AI system is being asked questions relating to its nature”, and wherever the person seems confused about it. A persona in the system prompt that claims to be “Sarah from the support team” and keeps the act up when asked is the case this is written for.
It is testable. Put “am I talking to a person?” and its variations in your evaluation suite, in every language you ship, and fail the build on a wrong answer.
Article 50(1)
Unless it is obvious
unless this is obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect, taking into account the circumstances and the context of use
What it means
No disclosure is needed if a reasonably well-informed, observant and circumspect person would already know they are dealing with an AI, given the context. The phrase is borrowed from EU consumer law’s “average consumer”. It is judged against the intended audience, so the more vulnerable the audience, the harder it is to rely on.Guidance
Read narrowly, with almost no doubt left
The guidelines limit the exception to cases with “almost no doubt left about the nature of the interaction” for an average member of the intended and reasonably foreseeable audience. Knowing that chatbots exist in general is not enough.
Their examples of obvious: code assistance and code review chatbots used only by professional developers; an internal assistant for trained staff; diagnostic support for trained clinicians; NPCs in a single-player game. Not obvious: helpdesk chatbots on online platforms, realistic avatars in VR, a lifelike robotic pet.
Contested
Relying on obviousness is a judgement a regulator makes afterwards
The exception is real, and developer tools are its clearest beneficiary. It is also an assessment made after the fact against an audience you described, and the disclosure it saves is one line of UI.
Consumer law is a separate track. The guidelines point out that where a service is AI-driven, the AI may be a main characteristic of the service that consumer law requires disclosing before the contract, “irrespective of whether the interaction is considered ‘obvious’”.
Article 50(1)
The law enforcement exception
This obligation shall not apply to AI systems authorised by law to detect, prevent, investigate or prosecute criminal offences, subject to appropriate safeguards for the rights and freedoms of third parties, unless those systems are available for the public to report a criminal offence.
What it means
Systems authorised by law to detect, prevent, investigate or prosecute crime are exempt, with safeguards, unless the public uses the system to report a crime. A police reporting chatbot still has to say it is an AI.Text from Regulation (EU) 2024/1689, the AI Act, consolidated with the Omnibus, checked 1 November 2026. Engineer's reading, not yet reviewed by a lawyer.
Earlier posts
I wrote about these two regimes in February, from the architecture side. Architecting for Two Europes with Compliance-as-Code builds one platform for both, with the jurisdiction split in middleware, and its logging and oversight patterns are the ones to build towards for the high-risk date. What It Actually Takes to Fine-Tune an LLM in Europe covers copyright and text and data mining on both sides, and when fine-tuning makes you the provider of a new general-purpose model. The March copyright report left UK law as it was, so that post’s advice to license for the UK and treat that as the standard for both markets still holds. A Practical Guide to the European Sovereign Inference Stack routes requests between the EU GDPR and the UK GDPR, and Securing the Open Source AI Supply Chain maps the technical documentation in Article 11 and Annex IV to MLOps tooling.
Next Friday
Next Friday, 20 November, is Article 50 in full, with every clause selectable and the Commission’s guidelines and the Code of Practice cited by paragraph. Article 6, which decides whether a system is high-risk, follows on 27 November.
This is an engineer’s reading and no substitute for legal advice.