Your AI Assistant Wants Your Workout Data. Here’s How It Plays Out.

Anthropic announced on Sunday (January 11, 2026) that Claude can now access Apple Health and Android Health Connect. This came just days after OpenAI announced ChatGPT Health on January 7, which also integrates with Apple Health and other wellness apps. Gemini already has access, as it’s owned by Google, which owns Android.

This means they can see every workout you’ve ever logged. Your runs with GPS routes. Cycling sessions with power data. Gym workouts with sets and reps. Swimming laps. Yoga sessions. All of it sits in Apple Health and Health Connect, and these AIs now have a way into it.

Granting Access: Of course, you need to grant access, but that won’t matter. The outcome will be the same. Here’s why.

These companies need all original data to train their models, be that web pages, digital TV programs, digital art…and your workout files and health records. They have to get it and, probably in reality, don’t care how they get it – that’s another subject (ask in the comments)

Here’s their likely playbook:

1. Now: “Why not connect your Apple Health and I’ll analyse your training!” They’ll tell you your weekly mileage, explain your heart rate zones, and maybe spot that you’re overtraining. Simple access, only working on your device. They promise they’re not training on it.
2. Soon: “do you want personalised training plans based on what actually works for athletes like you? Here’s the opt in button.” The AI will need workout data from millions of users to know that runners in your age group who overcook Sunday long runs fully recover by Wednesday. Or that cyclists of your age and ability have the biggest FTP gains with specific interval patterns. The features will be legitimately useful, proven by data, so people will opt in. (You might not, but they all will)
3. Endgame: These AI companies end up with a workout database that makes Strava’s 100 million current users look small (ChatGPT currently 800m weekly, perhaps with 20% saving workouts to their apps). The eventually dominant AIs will have larger active user bases. If even 10% connect their workout data, that will overtake Strava.

And it’s not just routes. It’s:

• Every run with pace, heart rate, elevation, and GPS track
• Cycling power data, cadence, and segments
• Swimming splits and stroke counts
• All cross-referenced with sleep, recovery metrics, and injury patterns
• Everything.

Remember when Strava’s heatmap accidentally revealed the locations of military bases? That was just from GPS tracks from a relatively small number of people who forgot their activities were flagged as public. Now imagine AI companies training on complete workout histories from hundreds of millions of users, knowing exactly where people train, when they train, how hard they train, and how their bodies respond. Strava should be worried. With ‘all’ the data and a superior intelligence model, it should be able to determine the ideal official segments, avoid duplication/overlap, and automatically keep cheaters and mis-recorded workouts off the leaderboard.

Listen to the discussion

Use the integrations if you want. Just know that “not currently training on your data” and “will never train on your data” are different sentences. When they ask for permission to “improve recommendations for everyone,” that’s when your Sunday workout becomes part of the dataset.

Your local running routes, your favourite training spots, your workout patterns—all of it could (will) end up in a database that makes Strava’s look trivial.

My Thoughts

I find what’s happening with AI and our data to be both deeply interesting and deeply troubling.

Perhaps I’m blinkered, but I can’t see this particular issue discussed here playing out in any other way (tell me why I’m wrong, I hope I am). Remember all those people who vowed never to be slaves to their mobile phones? They’re the ones sitting next to you right now, glued to it, longing to get home to be glued to Netflix on their other screen. AI is their next digital master, maybe yours too – this time it’s AI software, not the phone or TV hardware.

I wrote an article earlier in the month about Strava’s IPO. I doubt the timing of the IPO filing had anything to do with these developments, though clearly (to me) they represent a massive longterm threat to Strava’s business model. Perhaps Strava’s solution is to eventually license a link to our data in the AI models rather than capturing it directly from each of us?

What do you think?

https://the5krunner.com/2026/01/09/strava-ipo-filing-3-billion-valuation-analysis/

Last Updated on 31 January 2026 by the5krunner

the5krunner

tfk is the founder and author of the5krunner, an independent endurance sports technology publication. With 20 years of hands-on testing of GPS watches and wearables, and competing in triathlons at an international age-group level, tfk provides in-depth expert analysis of fitness technology for serious athletes and endurance sport competitors.