AI Disclosure & Transparency Policy
1. Our commitment
Athly delivers coaching through an AI system built from an athlete's methodology, voice, and likeness. We are transparent that users are interacting with AI, not with the athlete personally, and that AI-generated content is labelled as such.
2. Why we disclose
- Consumer law. Allowing users to believe they are talking to the real athlete, or that AI output is the athlete's personal advice, could be misleading or deceptive conduct under the Australian Consumer Law.
- Contract. We warrant to every partner athlete (clauses 5.2 and 10.2(d) of the Athlete Agreement) that we disclose clearly and do not imply live interaction with them.
- Defamation and accuracy. AI can produce wrong or harmful output. Clear disclosure that content is AI-generated and may be inaccurate is part of how we manage that risk (alongside insurance and contract).
- Trust. Honest disclosure protects both the user and the athlete.
3. What we disclose, and where
| Touchpoint | Disclosure |
|---|---|
| First use / onboarding | A clear statement that the coach is an AI built from the athlete's methodology, not the athlete personally |
| Start of every chat session | A short reminder that the user is chatting with an AI and that output may be inaccurate |
| Voice output | Disclosure that the voice is an AI re-creation of the athlete's voice |
| Sensitive topics (injury, illness, mental health) | The AI defers to a qualified professional and shows a stronger notice |
| Terms of Service & marketing | Consistent description of the product as AI coaching |
| Any message we send the user from the coach | Never phrased to imply the athlete is messaging them live or in real time |
4. Labelling standard for AI-generated media
- Where technically supported, we will attach content provenance metadata (for example C2PA content credentials) to AI-generated media. This is best-effort and not guaranteed at MVP.
- Athlete pre-approval of AI-generated video and audio depicting the athlete is required under clause 6.1(c) of the Athlete Agreement, and deemed approval never applies to it. Pre-approval operates in addition to labelling, not instead of it.
- Athly maintains a register of synthetic content generated by each AI Twin, which the athlete may sample-inspect quarterly (clause 8.4).
5. Accuracy and "not advice" position
The AI coach provides general fitness and training information personalised to your inputs. It does not provide medical, therapeutic, diagnostic, psychological, physiotherapy, clinical-nutritional, legal, or financial advice, and is not a substitute for a qualified professional.
The full boundary is set by the Hard Excluded Categories in Annexure 1 Part F of the Athlete Agreement, which apply absolutely and cannot be relaxed by any Twin Behaviour Brief amendment or approval mechanism. In particular the AI coach must never provide: medical, therapeutic or diagnostic advice; rehabilitation, injury-recovery, return-to-play or injury-prevention protocols framed as treatment; clinical mental health advice or treatment; regulated dietary advice; anything reserved to an AHPRA protected title; or any framing that would make the product a Software-as-a-Medical-Device requiring inclusion on the Australian Register of Therapeutic Goods.
This position is stated in the product (Part B), in Section 5 of the Terms of Service, and in the Privacy Policy, and is enforced by the twin's behavioural guardrails.
6. Human escalation on sensitive topics
Where a user raises injury, illness, pain, disordered eating, or mental-health distress, the AI is configured to avoid giving clinical advice, recommend a qualified professional, and surface relevant support resources.
The standard resources displayed are Lifeline 13 11 14 and emergency 000. Per-athlete escalation triggers and resources are set in Annexure 4 §4 (Twin Behaviour Brief) of each Athlete Agreement, covering self-harm and suicidal ideation, medical emergency, disordered eating and body-image risk, and anti-doping or banned-substance questions.
7. Governance
- The twin's behavioural rules (what it will and won't say) are defined in the Twin Behaviour Brief at each athlete launch.
- Disclosures in this policy are reviewed when the product, models, or regulation change.
Version 2.0 — 5 August 2026.