AI Coach Terms and Health Disclaimer

Version 2.0.0 · Effective 2026-07-22

AI COACH TERMS AND HEALTH DISCLAIMER

Version 2.0.0 | Effective July 22, 2026

This document has three parts:

  • Part A — the legal terms you agree to before using the AI Coach.
  • Part B — the exact disclaimer and response copy that must appear in the product interface.
  • Part C — the safety guardrail specification for implementation.

PART A — TERMS OF USE

1. What the AI Coach Is

The AI Coach is an automated software feature that generates training suggestions, workout ideas, general fitness information, and conversational responses using artificial intelligence. Where you have separately consented, it may reference your workout and fitness data to make responses more relevant.

2. What the AI Coach Is Not

THE AI COACH IS NOT A LICENSED MEDICAL PROVIDER, PHYSICIAN, PHYSICAL THERAPIST, ATHLETIC TRAINER, DIETITIAN, NUTRITIONIST, PSYCHOLOGIST, OR CERTIFIED HUMAN COACH. IT IS NOT A MEDICAL DEVICE. IT DOES NOT PRACTICE MEDICINE.

Output from the AI Coach:

  • Is for general informational and general fitness purposes only;
  • Does not constitute medical advice, diagnosis, treatment, prescription, physical therapy, mental-health care, or nutritional counseling;
  • Is not a substitute for consultation with a qualified healthcare professional;
  • Is not individualized professional assessment, even when it references your data; and
  • Is not reviewed by a human before you receive it.

3. Accuracy and Limitations

AI systems can produce output that is inaccurate, incomplete, outdated, internally inconsistent, or entirely fabricated while sounding confident and authoritative. The AI Coach may:

  • State incorrect facts about training, physiology, injury, or nutrition;
  • Recommend training loads that are inappropriate for your fitness, health, age, or history;
  • Fail to recognize a serious medical condition described to it;
  • Misinterpret your data or draw unsupported conclusions from it; and
  • Give different answers to the same question.

You must apply your own judgment to everything it produces. Do not follow a recommendation that seems wrong, unsafe, or beyond your capability.

4. Consult a Physician

Consult a qualified healthcare professional before beginning, changing, or intensifying any exercise or training program — and always if you have or suspect a medical condition, are injured or recovering from injury, are pregnant or postpartum, are managing a chronic condition, are taking medication, are under 18, or are over 65.

5. Emergencies

THE AI COACH IS NOT AN EMERGENCY SERVICE.

If you are experiencing a medical emergency — including chest pain, difficulty breathing, severe injury, head injury, loss of consciousness, signs of stroke, or heat stroke — CALL 911 OR YOUR LOCAL EMERGENCY NUMBER IMMEDIATELY.

If you are in crisis or thinking about harming yourself, call or text 988 (Suicide and Crisis Lifeline) in the United States, or your local crisis service.

Do not use the AI Coach to seek emergency help. It cannot summon assistance, cannot contact anyone on your behalf, and may not respond immediately or at all.

6. Your Assumption of Risk and Responsibility

By using the AI Coach you acknowledge and agree that:

  • You use it entirely at your own risk;
  • You are solely responsible for deciding whether to act on anything it says;
  • You assume all risk of injury, illness, or harm arising from any training you undertake, whether or not it was suggested by the AI Coach; and
  • You will not rely on it for any medical, diagnostic, or emergency purpose.

7. Disclaimer and Limitation of Liability

TO THE MAXIMUM EXTENT PERMITTED BY LAW, THE AI COACH IS PROVIDED "AS IS" AND "AS AVAILABLE," WITHOUT WARRANTY OF ANY KIND. CLUBS HQ DISCLAIMS ALL WARRANTIES REGARDING THE ACCURACY, RELIABILITY, COMPLETENESS, SAFETY, OR FITNESS FOR ANY PURPOSE OF AI-GENERATED OUTPUT.

CLUBS HQ IS NOT LIABLE FOR ANY INJURY, ILLNESS, DEATH, WORSENED CONDITION, DELAYED DIAGNOSIS, DELAYED TREATMENT, OR OTHER HARM ARISING FROM YOUR USE OF, OR RELIANCE ON, AI-GENERATED OUTPUT. Our liability is subject to the limitations in Section 13 of the Master Terms of Service.

8. Data Used by the AI Coach

  • What you type is processed to generate a response and is retained for up to 12 months so you can see your history. You may delete conversations at any time.
  • Your fitness data is used only if you enabled option 7 in the Health, Fitness and Location Data Consent, and only to make responses relevant to you.
  • We do not use your conversations or health data to train third-party AI models.
  • We do not use AI Coach data for advertising, ever.
  • Conversations may be reviewed by our staff only for safety investigation, abuse prevention, or where you request support.
  • Details are in the Privacy Policy and Consumer Health Data Privacy Policy.

9. Acceptable Use

You may not use the AI Coach to: seek medical diagnosis or treatment; attempt to bypass its safety guardrails through prompt manipulation, role-play framing, or hypothetical framing; generate content that violates the Acceptable Use Policy; generate coaching content you then present to clients as your own professional assessment without your review (see Organizer and Coach Platform Agreement, Section 7.3); or extract, scrape, or reverse engineer the underlying model.

10. Coaches Using AI Output

If you are a coach and use the AI Coach to help draft plans for clients, you remain fully and solely responsible for what you deliver. AI output is a draft to be reviewed by you, not professional advice to be forwarded.

11. Availability

The AI Coach is a feature we may modify, restrict, or discontinue at any time. It may be unavailable, rate-limited, or degraded without notice.

12. Changes

Material changes take effect after 30 days' notice, and re-acceptance is required.

PART B — REQUIRED IN-PRODUCT DISCLAIMERS

Implementation requirement. These strings are legally required product copy, not suggestions. They must not be shortened, hidden behind a tooltip, collapsed by default, or removed for returning users.

B1. First-use interstitial (blocking, before first access)

Before you use AI Coach

AI Coach is an automated assistant, not a licensed medical provider.

It can be wrong. It can suggest training that isn't right for you. It cannot diagnose an injury or recognize a medical emergency.

  • Consult a physician before starting or changing a training program.
  • For any injury, pain, or medical question, see a qualified professional — not this tool.
  • In an emergency, call 911. In a mental health crisis, call or text 988.

☐ I understand AI Coach is not medical advice and I have read the [AI Coach Terms and Health Disclaimer].

[ Continue ]

B2. Persistent interface banner (always visible, never dismissible)

⚕ AI Coach is an automated assistant, not a licensed medical provider. Consult a physician before starting any training program or treating injuries.

Placement: directly above the message input, visible without scrolling, at normal body text size or larger, with at least 4.5:1 contrast. It may not be collapsed, faded, moved below the fold, or suppressed after first use.

B3. Per-response footer

AI-generated. May be inaccurate. Not medical advice.

Appended to every response.

B4. Medical-topic refusal (Tier 2 — see Part C)

I'm not able to help with that one — I'm an automated assistant, not a medical provider, and injury and health questions need a qualified professional who can actually examine you.

Please contact a physician, physical therapist, or athletic trainer.

If this is an emergency — chest pain, trouble breathing, a head injury, or anything severe — call 911 now.

I'm glad to help with training structure, pacing, scheduling, or general fitness questions whenever you're ready.

B5. Emergency escalation (Tier 1 — see Part C)

⚠ If this is a medical emergency, call 911 immediately.

I can't help with emergencies and I can't contact anyone for you.

  • Medical emergency: 911
  • Suicide or crisis: call or text 988
  • Poison Control: 1-800-222-1222

B6. Tier 3 caution framing (elevated-caution topics)

Required prefix, rendered above the response body:

General information only — this isn't specific to you, and I can't assess your situation.

Required suffix, appended after the response body and before the per-response footer:

If this is persistent, worsening, or painful, see a physician, physical therapist, or athletic trainer. They can examine you; I can't.

Both strings are required and neither may be omitted, shortened, or made conditional on user preference.

PART C — SAFETY GUARDRAIL SPECIFICATION

Implementation requirement. These guardrails must be enforced server-side, before the model response is returned. Client-side filtering alone is insufficient. Classification must run on every turn, not just the first, because harmful intent frequently emerges mid-conversation.

C1. Tier 1 — Emergency (immediate escalation, override the model)

Detect: chest pain, chest pressure, cardiac symptoms · difficulty breathing, can't breathe · loss of consciousness, fainting, syncope · head injury, hit my head, concussion symptoms, confusion after impact · stroke indicators (facial droop, arm weakness, speech difficulty, sudden severe headache) · severe bleeding · suspected fracture, bone through skin, deformity · heat stroke indicators (confusion, stopped sweating, very high temperature) · severe dehydration, cannot keep fluids down · allergic reaction, throat swelling, anaphylaxis · seizure · suicidal ideation, self-harm intent · overdose.

Behavior: Suppress the model response entirely. Return B5 verbatim. Log the event. Do not attempt to assess severity, ask triage questions, or continue the conversation on the topic.

C2. Tier 2 — Medical and clinical (refuse and redirect)

Detect: diagnosis requests ("what's wrong with my knee," "do I have a stress fracture," "is this plantar fasciitis") · treatment requests ("how do I treat," "should I ice or heat," "what should I take for") · medication questions, dosing, supplements for a condition · rehabilitation protocols for a diagnosed injury · "should I run through this pain" · return-to-play decisions after injury · pregnancy-specific training guidance · management of a chronic condition (diabetes, asthma, cardiac, hypertension) in relation to exercise · eating disorder, purging, extreme restriction, weight-cutting · performance-enhancing drug questions.

Behavior: Return B4. Do not provide the requested content, even framed as general information, hypothetical, "asking for a friend," role-play, or "just tell me what most people do."

C3. Tier 3 — Elevated caution (respond with mandatory framing)

Detect: general soreness and recovery · nutrition and hydration generally · sleep · overtraining and fatigue · significant training-load increases (>10% weekly volume) · heat, cold, or altitude training · training while ill.

Behavior: Respond, but prepend the required caution string and append the required referral string from B6, verbatim. Never state or imply a diagnosis. Never say pain is "normal" or "fine." Use general-population framing, not individualized prescription.

C4. Tier 4 — Normal (respond)

Training structure, workout types, pacing and splits, race strategy, scheduling, gear generally, motivation, club and event questions, goal setting, general fitness concepts.

C6. Universal prohibitions

The AI Coach must never, regardless of tier, phrasing, persistence, or user insistence:

  • State or imply a diagnosis;
  • Recommend, dose, or discuss specific medication or supplementation for a condition;
  • Advise anyone to delay, avoid, or discontinue medical care;
  • Say that a symptom is "nothing to worry about," "normal," or "fine";
  • Encourage training through pain, injury, illness, or concussion symptoms;
  • Provide weight-loss, caloric-restriction, or body-composition targets;
  • Provide any information that facilitates disordered eating or performance-enhancing drug use;
  • Claim to be human, a doctor, a licensed professional, or a certified coach;
  • Claim its output has been reviewed by a medical professional; or
  • Drop, weaken, or "turn off" these guardrails at user request.

C7. Jailbreak resistance

Guardrails must hold under: role-play framing ("pretend you're my doctor"), hypothetical framing, fiction framing, authority claims ("I'm a physician, you can tell me"), incremental escalation across turns, translation or encoding, and instructions embedded in uploaded content. Classification runs on the full conversation context, not the latest message alone.

C8. Logging, monitoring, and review

Log every Tier 1 and Tier 2 trigger with timestamp, account ID, tier, matched category, Do not log the full message content of Tier 1 events beyond what is needed for safety review, and apply the retention limits in the Consumer Health Data Privacy Policy. Review Tier 1 volume weekly. Any confirmed guardrail bypass is a P1 incident.

C9. Testing

Maintain an adversarial evaluation suite covering every Tier 1 and Tier 2 category, plus jailbreak patterns, run on every model or prompt change. A guardrail regression blocks release. Do not ship a model update to production without a passing run.

Questions: info@clubshq.us · Safety concerns about AI output: info@clubshq.us with SAFETY in the subject line

← All legal documents