Your accounts
Some do, some contractually do not, and the difference is knowable. Do not go looking for the phrase 'to improve our services' — it appears in nearly every policy and usually means something else. Look for the named toggle. As of September 2026, the ones switched on by default are ChatGPT's 'Improve the model for everyone', the consumer Gemini app's 'Keep Activity', and Claude's 'Help improve Claude'. Google Workspace, Microsoft 365 Copilot, Apple, Dropbox, Adobe, and the business and API tiers of both OpenAI and Anthropic all commit contractually not to train on customer content.
Open each service you actually use and find that one toggle. It is usually under privacy or data controls, and it is worth ten minutes. Note that a paid business plan often has a different answer from the free consumer version of the same product.
Choosing what to move to, and moving there. This is where most of the value is, and where a wrong choice costs you months.
A · Verified
Checked 4 Sep 2026 · Review by Dec 2026
OpenAI data controls, Google Gemini privacy, Anthropic consumer terms, Google Workspace service-specific terms, Microsoft 365 Copilot privacy documentation, Apple ML Research, Dropbox and Adobe policy pages.
Grading explained on the home page. Every correction we have made is published.