How to stop Headspace training AI on your data
What the policy itself says, plan by plan. Where Headspace publishes the steps, they're here in its own words, read from an archived copy of the page that gives them. Where it doesn't, this page says so instead of inventing a menu path.
All accounts (free and Plus)
Headspace holds what a person does at 3am and what they tell a mental- health app about their mood, which makes the AI section worth reading closely. It is more careful than most: AI is defined, its uses are separated into behind-the-scenes work and features you interact with, and the policy promises to tell you before you use one and let you decline. None of that is about inputs. The quoted line puts "machine learning algorithms" among the purposes personal information is used for, and no form of the word "train" appears anywhere in 49,000 characters. Powering a model and training one are different claims, and only the first is made. The AI Features paragraph ends by pointing at a separate page on Headspace's approach to AI, which this site has now captured. It is a statement of philosophy rather than of data use — "We believe AI will not solve the world's most pressing problems by training only on web data" — and it describes Ebb, the AI companion, as "a complex AI system that contains multiple models". Its one operational sentence is about people: staff "review a subset of random Ebb conversations daily for quality assurance and continuous improvement". That is human review of what members tell the companion, in the service of improving it, and still not a statement that those conversations train a model. The pointer has been followed, and the question is open at the far end of it too.
Did a step here not work, or does Headspace offer a control this page doesn't mention? Tell us and we'll correct it.