NEWS

Be part of the Asia Pacific

The AI Regulation That Started on 2 August

A ring of stars over a motor yacht on calm night water

The EU AI Act’s transparency duties have applied since 2 August. They reach businesses a long way from Europe, because the test follows the output rather than the client. What they ask of a superyacht business in the Asia-Pacific, and what the Commission’s own guidance says you can leave alone.

There is an AI version of me. She has my face and my voice, and she says what I write for her. If this article had arrived as a video from her rather than as text from me, I would have to tell you so. That is the premise of the rule that came into force on 2 August: someone dealing with your business should be able to tell when they are dealing with a machine.

I recorded a briefing on it on 5 August and put it out to our industry, and in preparing it I went through the European Commission’s guidance paragraph by paragraph. The useful discovery was how much of ordinary daily practice that guidance leaves alone.

Does it reach you

The Act reaches a business through three doors: established in the EU, placing an AI system on the EU market, or the output of an AI system being used in the Union. The third door catches businesses based nowhere near Europe.

Your client’s nationality is not the test. The test follows the output. Use an AI assistant to draft the quotation you send to a yard in Italy and the output is being used in the Union. Same for listing copy read by buyers in Europe, and for an assistant that European crew and owners deal with directly.

For most APSA members the practical doors are a Med season, a European owner or yard, and AI-written material that reaches people in Europe. Ship operations and seafaring sit in scope alongside yachts.

What the transparency duty asks

Three things. Someone dealing with an AI system acting for your business should be able to tell it is AI, unless that is obvious. AI-generated or AI-altered images, audio and video should be identifiable as AI-made, and depictions of real people are disclosed by default. AI-written text published to inform the public on matters of public interest carries a disclosure, unless a named person has reviewed it and holds editorial responsibility for it.

Two points about that third one, because it is the one people misread. The disclosure sits where the dealing happens: a line in a terms page does not satisfy it, and neither does calling the thing an “assistant” without saying whose assistant it is. And routine sign-off does not reach the editorial bar. The named person has to be answerable for what went out.

Fines for breaching these requirements reach EUR 15 million or 3 per cent of worldwide annual turnover, whichever is higher, and for small and medium businesses the calculation reverses to whichever is lower.

Two questions settle most cases

Did the AI make the substance, or polish it? And is it public, or private?

The Commission’s guidance treats grammar correction, spellchecking, formatting, translation and readability edits as standard editing, at paragraph 90. None of that needs marking. An AI-written summary or a substantive rewrite is a different case, and does.

Private communications and internal documents sit outside the published-text duty. A quotation to a client, a report to an owner, a note to your team: outside. Your website, your newsletter, an article you contribute to a publication like this one: inside, unless you take the editorial route.

Images have their own answer, at paragraph 116. Colour correction, removing a passerby, lighting and noise adjustments, cosmetic touch-ups, and background replacement for aesthetic purposes in product advertising are all treated as minor. The duty bites where the editing changes how a person would assess the content’s authenticity. Swapping the sky behind the yacht sits on the minor side. Changing how the yacht itself appears does not.

For images and video the route is the mark. The editorial carve-out applies to text.

The EU AI Act briefing, a recorded briefing for maritime operators
The full briefing, recorded 5 August 2026.

Where this shows up in a superyacht business

Four places, going by the operations I work in.

The enquiry or charter assistant on your website, including anything a vendor embedded in a booking or contact page. Open it the way a client meets it and see whether it says what it is.

Listing and marketing imagery. Ask your agency what was changed, and whether the change alters how the yacht reads.

Published copy: the site, the newsletter, the industry article. Name the person who approves it.

Anything a vendor updated recently. Vendor updates change what a tool discloses, so this is a recurring check rather than a one-off.

The duty that has been law longer

Since 2 February 2025 the Act has required organisations covered by it to see that staff and contractors using AI on their behalf have a sufficient level of AI literacy. In July 2026 the wording softened: the duty is now to take measures to support the development of AI literacy, and no business has to guarantee any particular level of ability in any individual. The duty itself stands.

The Digital Omnibus on AI, in force since 27 July 2026, did delay part of the Act: the high-risk obligations, including AI used in recruitment and crew selection, moved to 2 December 2027, and AI embedded in regulated products to 2 August 2028. Transparency and AI literacy were left where they were.

No fine attaches to the literacy duty. Its weight is evidentiary. The training record is what an insurer, a client or a regulator asks to see, and its absence makes any other breach read worse. My own professional indemnity renewal asked about AI use this year, which is where a lot of people will first meet this.

Who carries it on a vessel

The Act does not answer this in as many words, and the answer matters more here than in most industries, because a yacht has an owning entity, a management company and a crew employment company all in the frame at once.

The duty sits with the deployer, meaning the organisation using AI under its own authority. Where crew use AI acting on behalf of the owning entity, the owning entity is the deployer for vessel operations and carries the duty. It does not fall on the crew member personally, and usually not on the crew employment company holding the seafarer agreement, because that company is not the one directing the use.

A management company is a deployer too, for the AI in its own office: the enquiry inbox, the accounts, the marketing. That is a separate register from the vessel’s, and keeping them separate is what makes either of them defensible.

Which gives you three registers to think about rather than one: the vessel’s, the management office’s, and the shoreside business’s if you run one. Each records who was trained, on which tool, when, in what format, who delivered it, when the refresher falls due, and the evidence attached. Superseded versions stay on file, because the history is part of what you are evidencing.

Three things to do this month

  1. Name the person who approves AI-assisted content before it is published, and write the name down. That is what makes the editorial route available to you for text.
  2. Open every customer-facing AI as a customer meets it, and check that it identifies itself. Re-check after vendor updates.
  3. Start the training register, even if the first row is you: who, on what tool, when, and the evidence attached.

An afternoon of writing things down covers most of an operation this size.

Dates to keep

DateWhat happens
2 February 2025AI literacy duty in force
2 August 2026Transparency duties apply
2 December 2026Marking grace period ends for tools already in use before 2 August 2026
2 December 2027High-risk duties, including AI used in recruitment and crew selection
2 August 2028AI embedded in regulated products

These dates changed in June and July 2026. Anything written before then may carry the old ones.

General information, not legal advice. Interpretation for your own business belongs with qualified counsel in your jurisdiction.

The full briefing, about twenty minutes, is at southernsky.ai/chart-room/eu-ai-act-briefing.


Kristina Agustin is the Founder and Principal Digital Navigator of Southern Sky AI, working with maritime and superyacht businesses on AI deployment, governance and training. Admitted Lawyer (Supreme Court of NSW). Southern Sky AI designed and built APSA’s website.

Scroll to Top