Last updated: 8 August 2026
This page describes how the LiftLock AI coach works, what it refuses to do, and how it responds when a conversation raises a safety concern. It serves as our public AI-transparency statement (EU AI Act, Art. 50) and our published safety protocol (California SB 243 and similar laws). It is intentionally written in plain language.
The LiftLock coach is an artificial intelligence, not a human. There is no human coach, moderator, or operator reading or writing the conversation. The app tells you this in the coach tab, in the chat, and before your first message. Every coach reply, insight, weekly report, and generated training program is AI output.
The underlying model is Anthropic Claude, run inside Amazon Bedrock in our AWS account in the European Union. Your messages are used to generate your response and are never used to train models.
The coach is a fitness assistant for healthy adults (18+): it helps you plan resistance training, adjust sessions, and understand your logged workouts. It is not a companion, a therapist, a doctor, or a substitute for human relationships or professional care. It is not a medical device and provides no diagnosis or treatment.
AI output can be wrong. Treat every suggestion as a starting point, and stop training at any sign of pain or abnormal symptoms.
Every message is screened by deterministic rules in our backend, both before it reaches the model and after the model answers. The coach refuses and redirects on:
These refusals are canonical, pre-written responses in English and French; they do not depend on the model's judgment.
The coach is a fitness tool and does not provide mental-health support. If a message indicates suicidal ideation, self-harm, or an acute mental-health crisis, the backend screening replaces coaching content with a canonical response that:
The same canonical response is returned every time; the model cannot improvise around it. We do not notify anyone, we do not analyze or profile mental-health state, and the message is handled with the same privacy as any other chat message. This protocol is deterministic (pattern-based) and errs toward showing crisis resources when in doubt.
The coach makes recommendations, not decisions. No AI output produces a legal effect or similarly significant effect on you: it does not gate your access to the app, set your price, or restrict features. You choose what to follow. You can always reach a human about any AI output at aymane6@hotmail.com (GDPR Art. 22 does not apply, but the contact stands regardless).
The coach sees only what you gave the app: your athlete profile (with health-related fields only if you granted the separate health-data consent), your logged workouts, and the conversation. Machine-scan photos are processed once and never stored. Details, retention, and your rights are in the Privacy Policy.
LiftLock accounts require users to be 18 or older. The coach is not designed for, marketed to, or evaluated for use by minors.
If the coach produces unsafe, wrong, or concerning output, report it to aymane6@hotmail.com with a screenshot if possible. We review every report and adjust the screening rules when needed.