What the disclosure duties from 2 August 2026 actually require, and how coaches can turn a legal floor
into a practice clients trust.

Colin Cosgrove
Founder, Movizmo Coaching Solutions · AI Governance Consultant · Leadership Coach · Board Member · August 2026

The Question I Keep Hearing

Coaches and L&D teams keep asking me the same question. When do I have to be transparent that I've used AI?

Article 50 of the EU AI Act has taken effect as of 2 August 2026. Transparency under the EU AI Act means a person's right to know, at the moment they encounter it, that they're dealing with AI, or that what they're seeing, hearing or reading was made or altered by AI. Breaches of Article 50 sit within the fine tier of up to €15 million or 3% of worldwide annual turnover under Article 99. This sits below the top tier reserved for prohibited AI practices, though it remains a substantial penalty (European Commission, 2024).

Article 50 sets out four disclosure situations. As 50(4) contains two distinct disclosure requirements, they're separated here for clarity. Each duty lands on a different party. Some sit with the coach as a deployer. Some sit with the technology company that built the tool. Knowing which is which is critical to protect you and your client.

What Article 50 Actually Says

50(1) AI interaction disclosure
Providers must design AI systems that interact directly with people so those people know they're dealing
with AI, unless this is already obvious to a reasonably well-informed, observant person given the context.
This is a provider obligation in the first instance. A coach using a third-party chatbot carries a narrower
practical duty: selecting a tool that already discloses AI status, and keeping that disclosure switched on. The exemption is narrow by design. The disclosure must be perceptible within the interaction itself; a statement buried in terms and conditions, a watermark alone, or a vague reference to an assistant falls short.

50(2) Synthetic content marking
Providers of systems that generate synthetic audio, image, video or text must ensure the outputs are
marked in a machine-readable format and detectable as artificially generated. This one sits almost entirely
with the tool builder. A coach who builds and deploys a custom project/agent or similar tool under their own name may themselves count as the provider for that tool.

50(3) Emotion recognition and biometric categorisation
Deployers of these systems must inform the people exposed to them.

50(4) Deepfake disclosure
Deployers must label content that resembles a real person, even without intent to deceive and even where
no real individual is depicted. Evidently artistic or satirical work still triggers disclosure, carried out in a
lighter form that leaves the work itself undisturbed.

50(4) AI-generated text on public-interest matters
Deployers publishing AI-generated or manipulated text meant to inform the public on public-interest
matters must disclose that origin, unless the text has gone through human review under someone's editorial responsibility. That exemption depends on a genuine, documented workflow with a named person holding editorial responsibility, a standard set out in Commission guidance interpreting the statutory text.
Provider and deployer are distinct legal roles, and most coaches sit on the deployer side. Article 50 sits
alongside other obligations. Meeting its requirements leaves separate duties under GDPR, consumer
protection law, and the high-risk regime elsewhere in the Act fully in place. Coaches or organisations working across borders should also note that territorial scope can extend beyond the EU in both directions.

What This Means for Coaches

Interaction disclosure (50(1))
This applies the moment an AI system talks to your client directly. An intake chatbot on your booking page qualifies. An AI coaching assistant your client uses between sessions qualifies. The trigger is whether the client is having a direct exchange with the system and might reasonably fail to notice. Using ChatGPT
yourself to prep a session plan sits outside this duty; your client never talks to the tool. A client chatting with a bot you've deployed sits inside it, and your role there is confirming the disclosure is present and switched on.

Synthetic content marking (50(2))
The machine-readable marking obligation belongs to the tool provider in almost every case. Where this
touches your practice is generated materials you hand to clients: an AI-written coaching plan, an AIgenerated image in a workbook. Ask your vendor whether marking is built in. Confirming it is your principal task here.

Deepfake and public-interest text (50(4))
For most one-to-one coaching, this carries low relevance. Your work centres on your client's own words and thinking, distinct from producing content that resembles someone else or informs public debate. The edge cases deserve a second look: a promotional video using an AI avatar of yourself, a marketing piece built from AI-generated audio. Step into content that resembles a real person, or publish material framed as informing the public on a public-interest topic, and this duty comes back into view.

Emotion recognition and biometric categorisation (50(3))
Also low relevance for most coaches. A tool reading vocal tone, facial expression or physiological signals
during a session, even as a minor feature, triggers this duty. Your client needs to know before that happens. Watch for this if you're availing of advanced notetakers that claim to read these signals.

“Thinking for yourself is the thing on which everything else depends.”
NANCY KLINE (1999)

Article 50 tells you when disclosure is legally required. Whether you are maintaining a strong human thinking partnership with your client is a deeper question. Our bar as coaches and learning professionals should sit significantly higher than what the law mandates.

For Organisations and L&D Teams

The provider and deployer split carries more weight once an organisation is involved. An L&D team that buys an AI coaching platform and rolls it out internally becomes the deployer. The disclosure duties in 50(1), 50(3) and 50(4) sit with that team regardless of what the vendor built into the underlying system.
Three practical moves follow. First, vendor due diligence should ask directly whether synthetic content
marking under 50(2) is genuinely built in, and request evidence to back that answer. Second, build disclosure practice into coaching supervision and programme design from the outset, alongside procurement decisions.

Third, name Article 50 duties directly in tool evaluation frameworks and internal AI policy, so the people
running sessions understand their obligations. This ties the whole exercise to governance work already
under way in most organisations.

Article 50 Compliance Checklist

Living the Transparency Value

Transparency here functions as the floor the law sets. Tell your client when you've used AI to prepare, in
every case. Write that same commitment into the agreement in language a client can actually picture. Offer a real choice: yes, no, or yes with limits. Revisit that choice periodically as the relationship continues. Hold AI-generated observations up for the client to examine together.

A mature coaching practice holds the following together: the legal floor Article 50 sets, the professional
standard a body like EMCC's code of ethics asks for around client welfare, competence and honesty, and a
coach's own personal answer to what transparency means with a client in the room. Treating all three as one connected practice keeps a coach ahead of the regulation, ahead of the professional standard, and in a position to earn the trust of their clients.

References

European Commission (2024) AI Act Service Desk: Article 50 – Transparency obligations for providers and
deployers of certain AI systems. Available at: https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50
(Accessed: 6 August 2026).
Kline, N. (1999) Time to Think: Listening to Ignite the Human Mind. London: Cassell Illustrated.

This paper was prepared with AI assistance. The analysis, framing and final content were directed and reviewed by the author. This paper offers professional commentary, not legal advice. It shouldn't be relied on as a substitute for advice by a suitably qualified professional on your particular circumstance. Interpretation of legal texts may vary.