I'm building Aevo, an endurance coaching app. An athlete connects Strava, and we use their activity data to coach that same athlete. Part of the coaching is generated with an LLM (OpenAI API), used for inference only.
I want to stay on the right side of the API Agreement, and I could not resolve this from the Agreement text or existing threads.
The Agreement is clear on two points: data obtained through the API may not be used to train AI or ML models, and a user's Strava data may only be displayed or disclosed to that same user. What it does not address is inference.
My question: is AI inference on an athlete's own data, used solely to generate guidance for that same athlete, permitted? To be specific about my setup:
- No model training, fine-tuning, or improvement on Strava data, ever.
- A user's data is only ever shown to that user, never to others, never aggregated.
- Zero data retention can be enabled on the OpenAI account so inputs are not stored.
- Only derived metrics (pace, distance, HR summaries) need to be sent, not raw activity streams.
- Real-time via webhooks, official Connect with Strava button, and every activity links back to Strava.
I found an earlier thread asking essentially the same thing that never got an official answer (developers-api-7/ai-inference-with-strava-data-is-it-prohibited-under-the-new-api-agreement-13256). A clear yes or no would help a lot of developers, not just me.
If inference in this form is not allowed, I would rather redesign now than keep guessing. Thanks.
