ホーム / ShortFlingブログ / Safety

この記事は現在、英語版のみ提供しています。英語版を読む →

Safety

EU AI Act Article 50: What It Means for AI Stories and Chat

ShortFling編集部·Nov 11, 2026·Updated Nov 11, 2026·5分で読めます

The short answer

Since August 2, 2026, Article 50 of the EU AI Act has required three things that matter for AI stories and characters. An AI system that talks directly with people must tell them they are talking to an AI unless that is obvious. Tools that generate text, images, audio or video must mark their output in a machine-readable way. And people who publish certain AI content, namely deepfakes and AI-written text informing the public on matters of public interest, must label it.

For most fiction writers using AI as an editor, the direct obligations are narrower than the headlines suggest. The machine-readable marking falls on the companies that build the tools, and the text labelling rule targets public-interest publications such as news, not novels. For AI chat characters and interactive story apps, the duty to disclose the AI is real and already in force. This post sticks to text and conversation; for video, see our guide to AI-generated video labeling rules. It is an overview as of mid-October 2026, not legal advice.

What is in force, and what was not delayed

The 2026 AI Omnibus, Regulation (EU) 2026/1744, which entered into force on July 27, 2026, postponed the Act’s high-risk rules to December 2, 2027 and later. According to a Goodwin client alert of August 3, 2026, it did not change the substance of the Article 50 transparency duties, which became applicable and enforceable by national authorities on August 2, 2026, apart from one narrow grace period for marking.

That grace period is limited. The European Commission’s official Q&A says providers of generative systems already on the market before August 2, 2026 have until December 2, 2026 to meet the machine-readable marking duty. The other Article 50 duties apply now. The same Q&A says content generated before August 2, 2026 does not need to be labelled retroactively, though the Commission encourages it where possible.

Fines for breaching these duties can reach €15 million or 3% of worldwide annual turnover, whichever is higher, under Article 99(4). Enforcement sits mainly with national market surveillance authorities, with the EU AI Office covering certain systems built on general-purpose models.

AI characters and chat: you must be told

Article 50(1) is the rule for chatbots, AI companions, role-play characters and interactive story apps. Providers must design these systems so people are informed they are interacting with an AI, unless that is obvious to a reasonably well-informed, observant and circumspect person in the circumstances. The notice must be clear and distinguishable and given at the latest at the first interaction.

The Commission’s guidance narrows who is covered and how. The duty applies to systems built for a genuine two-way exchange in which the AI communicates directly with a person. The Commission says the obvious-from-context exception should be read restrictively, because it deprives people of transparency. A platform that sells itself as an AI character app may argue its users already know; a character designed to feel human, inside a product that blurs the line, is where that argument gets weakest.

Users often ask whether the rule outlaws a character who stays in role. On its face, Article 50 does not ban role-play; it requires the provider to inform users at the latest at the first interaction. How much more is needed when a character later claims to be human is a question for the Commission’s guidelines and, in the end, national regulators. Separate rules, such as California’s companion chatbot law covered in our post on SB 1119, add their own requirements outside the EU.

Does AI-written fiction need a label?

Article 50(4) has two parts. The first covers deepfakes: image, audio or video that resembles real people, places or events and would falsely appear authentic. Text is not part of that definition. For evidently artistic, creative or fictional works, the deepfake disclosure is limited to a form that does not hamper enjoyment of the work.

The second part covers text, and it is narrow. Deployers must label AI-generated or manipulated text published to inform the public on matters of public interest, which the Commission’s Q&A lists as subjects like politics, public health, justice, consumer safety and economic or scientific developments. A romance serial or fantasy novel does not obviously fit that description. Even for covered text, the label is not required where there has been human review or editorial control and someone holds editorial responsibility.

There is also a personal-use carve-out in the definition of deployer: someone using AI in a personal, non-professional activity is not a deployer under the Act. Platform and retailer rules are a separate matter. Amazon’s publishing platform, for example, has its own AI disclosure questions, covered in our post on KDP AI disclosure.

Marking is the toolmaker’s job

Article 50(2) requires providers of systems that generate synthetic text, images, audio or video to mark outputs in a machine-readable, detectable format, as far as technically feasible. It does not apply where the AI performs an assistive function for standard editing or does not substantially alter the input, and the Commission’s guidelines give examples of what counts as standard editing.

To help companies comply, the Commission backs a voluntary Code of Practice on Transparency of AI-generated Content. The Commission and the AI Board have confirmed it as an adequate tool for demonstrating compliance, and the Commission said about 190 organisations had signed it by the end of July 2026. For writers, the practical upshot is that marking, where required, should come from the tool rather than from you.

What this means for readers and writers

If you use AI chat characters in the EU, you should now be told clearly, at the start, that you are dealing with an AI. If an app does not, that is a question for the provider. If you write fiction with AI help, Article 50 most likely does not require you to put a label on the story itself, but honesty with readers is still good practice, and some platforms and stores require disclosure under their own rules. Our post on AI slop books and reader trust covers why that matters commercially.

Where ShortFling fits

ShortFling is a reading and writing platform, not an AI chat or companion app, and we are not making compliance claims in this post. In the writing studio, the AI editor is presented as an AI tool by name, it is free within a daily allowance, and you remain the author: work must be your own original writing with adult characters only, and pressing Publish submits it for review, so it goes public only after approval.

Finished stories live in the fiction library, with separate SFW and 18+ NSFW sections; NSFW availability depends on your region and platform. This is a plain-language overview, not legal advice.

FAQ

When did EU AI Act Article 50 start to apply?

On August 2, 2026. The 2026 AI Omnibus did not delay it as a whole. The only grace period is for machine-readable marking by providers of generative systems already on the market before that date, who have until December 2, 2026.

Do AI chatbots and characters have to say they are AI in the EU?

Yes, under Article 50(1), providers must inform people that they are interacting with an AI system at the latest at the first interaction, unless it is obvious to a reasonably well-informed person. The Commission says that exception should be read restrictively.

Do I have to label a novel I wrote with AI help?

Probably not under Article 50. Its text labelling duty covers AI text published to inform the public on matters of public interest, and it does not apply where there is human review or editorial responsibility. Stores and platforms may have their own rules. This is an overview, not legal advice.

Does AI content made before August 2026 need a label?

No. The European Commission’s Q&A says content generated before August 2, 2026 does not need to be labelled retroactively, although it encourages labelling where possible.

Sources

  1. EU AI Act, Article 50: Transparency obligations (full text) (checked 2026-10)
  2. European Commission: Q&A on transparency obligations under Article 50 (grace period, no retroactive labelling, public-interest text, enforcement) (checked 2026-10)
  3. European Commission: Guidelines on transparency obligations for providers and deployers of AI systems (checked 2026-10)
  4. European Commission: Code of Practice on Transparency of AI-generated Content (about 190 signatories by end of July 2026) (checked 2026-10)
  5. Goodwin: Not delayed, not deferred, EU AI Act transparency obligations now in force (Aug 3, 2026; AI Omnibus, fines under Article 99(4)) (checked 2026-10)

ShortFlingで近日公開

最初の縦型ドラマが近日公開。各シリーズの第1話は無料になります。アプリは不要です。

▶ ラインナップを見る

読み続ける

こちらもおすすめ