May 27, 2026
Your data, their glow-up?
PostHog will train AI models with your data (opted-in by default)
PostHog says your data will help its AI unless you switch it off — and users are not buying the wording
TLDR: PostHog says it wants to use customer data to make its AI tools smarter, with most U.S. users included unless they turn it off. Commenters fixated on both privacy concerns and the company’s wording, roasting "opt-in by default" as a slick way of saying opt-out.
PostHog wanted to pitch a shiny future: smarter tools, automatic bug spotting, and even AI that could predict when customers might get confused before a product goes live. But the internet immediately grabbed the real headline by the collar: your data may be used to train those systems unless you opt out. That one detail turned the announcement into a full-on comment section roast.
The loudest reaction was simple: this feels like a betrayal. One user said they picked PostHog as a more private alternative to Google Analytics, only to now watch it drift toward what they called "enshitification." Ouch. Others were less focused on privacy panic and more obsessed with the wording, turning the thread into a grammar fight with torches. PostHog described the plan as "opted-in by default," and commenters pounced. "Opt-in by default = opt-out?" one person deadpanned, while others called it an outright oxymoron. The vibe was basically: you cannot call it a choice if the choice was already made for me.
To PostHog’s credit, the company did lay out the rules in public instead of sneaking them into fine print: European Union users are off by default, U.S. cloud users are on by default, data will be anonymized, and training starts June 29. But in the court of public opinion, the technical details got steamrolled by the semantics war. The community has spoken, and apparently the hottest AI debate of the week is sponsored by the phrase "that’s not what opt-in means".
Key Points
- •PostHog says it wants to train AI models on data stored in customer PostHog instances to improve existing AI features and build new products such as PostHog Code.
- •The company identifies session replay analysis as its first target area, saying trained models could make replay-based issue detection cheaper and more scalable.
- •PostHog is also exploring synthetic user testing and behavior prediction to detect potential user confusion, identify broken flows, and suggest product changes.
- •According to the article, EU cloud users and customers with agreements that prohibit training are excluded by default, while most US cloud users are included by default unless they opt out.
- •PostHog says it will anonymize data, use only data already in a customer's instance, perform the training itself, avoid third-party model providers, and delay the start of training until June 29.