Stack Overflow CEO says the company plans to charge large AI developers for access to the 50M questions and answers on its service as soon as mid-2023
The programmer Q&A site joins Reddit in demanding compensation when its data is used to train algorithms and ChatGPT-style bots Mastodon: @paul@social.paulkedrosky.com . Tweets: @nwaisb , @jason , @evijitghosh , @abacaj , @oliverjumpertz , @sarthakgh , @vivek7ue , @peard33 , @technologypoet , and @ruchowdh . Thanks: @tsimonite Mastodon: Paul Kedrosky / @paul@social.paulkedrosky.com : More and more companies belatedly realizing they own valuable AI training sets, and that they have had them looted. Stack Overflow joins the list of companies newly charging for AI training access. … Tweets: Noah Waisberg / @nwaisb : LLMs raise sooooo many interesting IP issues. One big one: were previous un-paid/un-permissioned training runs done on sites including these breach of copyright or breach of contract (the contract being the website terms and conditions)? https://twitter.com/... @jason : The list of data holders who will require AI products to license datasets grows: Reddit, Twitter, & now Stackoverflow https://www.wired.com/... Avijit Ghosh / @evijitghosh : This money is coming back to users who create the actual content. Right? RIGHT? https://www.wired.com/... Anton / @abacaj : Looks like the data moats are now making moves to defend that https://twitter.com/... Oliver Jumpertz / @oliverjumpertz : I said it first, and now it has happened. You just cannot expect to scrape data from all over the internet to build LLMs without compensating the platforms and people who created that content in the first place. https://www.wired.com/... Sar Haribhakti / @sarthakgh : StackOverflow and Reddit want to charge for their UGC content Quora wants to own the consumer interface by working with multiple model providers https://www.wired.com/... Vivek Raghunathan / @vivek7ue : Another shoe drops ... https://www.wired.com/... Wonder how Reddit and Stack will draw the line between who pays and who doesn't? What about open-source LLM efforts like @AiEleuther GPT-J/GPT-Neo or @StanfordCRFM Alpaca or @FacebookAI Llama / OPT or @databricks Dolly? Paresh Dave / @peard33 : Stack Overflow has 50 million questions and answers related to computer programming on its service. They are all available free for others to use under a Creative Commons license. But Stack's CEO says big tech companies are violating the license https://www.wired.com/... Vanessa Harris / @technologypoet : Subscription fees will save us from super advanced AI. https://twitter.com/... @ruchowdh : Seeing a trend now between Reddit and Stack Overflow - it seems the only party NOT cashing in on tech billions are users - you know, the actual content generators. https://www.wired.com/... Thanks: @tsimonite
Context & Ripple Effects
Stack Overflow is the second major user-generated content platform, after Reddit's $60M-a-year data deal, to declare that AI training on its corpus should be a paid license rather than free extraction. The CEO's framing — that 50 million programming questions and answers were effectively 'looted' by ChatGPT-style bots — echoes the publisher grievance raised in the writers-and-publishers call to band together against OpenAI and peers.
The move matters because code Q&A is among the highest-value training data there is: it pairs natural language with verified answers, exactly what models like ChatGPT are built on. What followed validates the threat — within a year, Stack Overflow converted the stance into an actual data supply deal with OpenAI, complete with product integrations.
First-order effects
- Large AI developers training on Stack Overflow's corpus face a new licensing bill from as soon as mid-2023, ending the era of free bulk access to its 50M questions and answers.
- Stack Overflow gains a potential new revenue line from data licensing at a moment when its core traffic is threatened by the very chatbots trained on its content.
Second-order effects
- Reddit's parallel stance hardened into market-rate benchmarks — reportedly $60M/year from Google and an OpenAI partnership — giving Stack Overflow and other forum operators a reference price for negotiating their own deals.
- AI labs that skip licensing now face enforcement risk rather than mere bad press: Reddit's later lawsuit against Anthropic over alleged post-cutoff scraping signals that unpaid access carries legal consequences.
Third-order effects
- If the pattern holds, community-contributed knowledge bases restructure from open commons to licensed data suppliers, with platform owners — not the contributing users — capturing the licensing value, and 'scrape first, pay never' replaced by negotiated API contracts as the default path for AI training data.
The trend: Owners of high-value human-generated corpora are converting free AI training access into paid licensing deals, with Reddit and Stack Overflow setting the template and litigation closing the loophole.