Stack Overflow users deleting answers after OpenAI partnership
Stack Overflow’s new licensing deal to let OpenAI train on its Q&A content has triggered a backlash from contributors, some of whom are deleting or sabotaging their answers in protest. Commenters argue over whether this violates the spirit or letter of Stack Overflow’s Creative Commons license, whether large language models should be allowed to ingest user-generated knowledge without individual attribution or compensation, and whether the platform is effectively monetizing unpaid labor. Others counter that the content was always permissively licensed, that mass deletions undermine the commons more than they hurt AI companies, and that Stack Overflow’s limits on removals are necessary to keep the site usable.
Reaction to the Stack Overflow–OpenAI Deal
- Many see the partnership as SO “selling” volunteer work to a for‑profit AI vendor without sharing revenue or credit.
- Some argue nothing fundamentally changed: content was always CC‑BY‑SA and monetized via ads/SEO; LLMs are just another consumer.
- Others say the deal breaks the “spirit” of the original bargain: answers were meant for humans and the open web, not paywalled AI products.
Licensing, Attribution, and Fair Use
- Repeated focus on CC‑BY‑SA:
- Attribution requirements are hard or impossible to honor in LLM outputs.
- ShareAlike raises questions about whether models or weights must be open‑licensed.
- Some claim LLM training is (or will be ruled) fair use, making CC terms irrelevant; others strongly dispute that.
- Confusion and debate over SO’s past license changes (3.0 → 4.0) and whether relicensing or exclusive data deals can be legally challenged.
User Protests: Deletions and Sabotage
- Some contributors delete or attempt to delete answers; SO reverses mass deletions and suspends accounts, a policy that predates the deal.
- Suggestions to “upvote bad answers” or slowly inject low‑quality/AI‑gibberish content to poison future training data; others call this unethical “scorched earth” that pollutes the commons and hurts users more than AI firms.
Incentives, Value, and “Unpaid Labor”
- Many feel like uncompensated labor for SO and OpenAI; loss of attribution and traffic undermines the motivations (helping others + recognition).
- Counterpoint: contributors were never promised payment; they traded labor for a useful free service and visibility, and that deal still holds formally.
LLMs vs Stack Overflow
- Some now prefer LLMs as a first stop: faster, more conversational, less “toxic” or gatekept than SO.
- Others find LLMs too error‑prone, especially on niche/edge cases, and still rely on archived SO threads and human discussion.
- Concern that LLMs, trained on SO, will siphon traffic and eventually kill the site, weakening future training data.
Moderation, Data Retention, and Rights
- Many criticize SO for blocking deletions of accepted/valuable posts; defenders liken it to wiki‑style preservation.
- Questions raised about GDPR “right to be forgotten”; unclear how far deletion rights extend beyond personal data.
Broader Platform and Power Issues
- Discussion of “technoserfs”: users generating monetizable datasets for platforms and AI firms.
- Some call for non‑profit or community‑governed Q&A alternatives; others doubt they can overcome network effects.