Does Stack Overflow train on your data?
Stack Overflow guarantees its AI providers won't train on what you type, then takes a perpetual, irrevocable licence to it for "product development" without saying whether that includes training.
This clause is real, but it doesn't resolve the training question - see the finding below.
"You retain full ownership of your AI Inputs, as applicable, but you grant to Stack Overflow a worldwide, perpetual, irrevocable, non-exclusive, sublicensable, right to use, copy, modify, distribute, and store the AI Inputs."
- 17 Aug 2026
- 5 Oct 2026
- Public Network (AI Features)
- global
- Fetched from the company
- 220F6BC7…D45F
Why this is still UNCLEAR: The AI Addendum is unusually explicit about third parties and silent about Stack Overflow. It requires that trusted third-party providers "do not use the input content or prompts for training their machine learning models" and repeats the same for outputs — a real commitment, plainly written. Against that, the licence quoted here is perpetual and irrevocable, and the purposes it names are "analytics, product development, and product improvement". Training is neither stated nor excluded, and a company that knew how to write the words "for training their machine learning models" one paragraph earlier did not write them here. That asymmetry is the finding. Note separately that public posts are CC BY-SA licensed and published as a data dump, so the training question for Subscriber Content is answered by that licence rather than by this addendum.
What each plan can refuse
The same questions, asked of every plan we could identify. Every cell comes from a quoted clause or a documented search - see the evidence above.
- Not established
- Not stated
- —
- —
- Not established
How to opt out of Stack Overflow AI training
Plan by plan, from what Stack Overflow's own policy says about refusing. Where it publishes the steps, they're listed here with the limits they carry. Where there's nothing to switch off, or nothing graded, this says so instead of inventing a menu path.
Public Network (AI Features)
Not graded, so no route is pointed at. The AI Addendum is unusually explicit about third parties and silent about Stack Overflow. It requires that trusted third-party providers "do not use the input content or prompts for training their machine learning models" and repeats the same for outputs — a real commitment, plainly written. Against that, the licence quoted here is perpetual and irrevocable, and the purposes it names are "analytics, product development, and product improvement". Training is neither stated nor excluded, and a company that knew how to write the words "for training their machine learning models" one paragraph earlier did not write them here. That asymmetry is the finding. Note separately that public posts are CC BY-SA licensed and published as a data dump, so the training question for Subscriber Content is answered by that licence rather than by this addendum.
What this policy has done since we started watching
No change detected. We've held a copy of Stack Overflow's documents since 13 Aug 2026 and re-read them since, across 3 captures of 2 documents. The text hasn't moved in a way that touches this verdict. That's a fact about the documents, and it's only knowable by having kept the earlier copies.
Vendor response
We haven't yet written to Stack Overflow about this verdict, and they haven't disputed, clarified or announced a change to any clause quoted above. When either happens it's published here, word for word.
Report a wrong clause