PostHog will train AI models with your data (opted-in by default)
5 hours ago
- #Data Privacy
- #Product Development
- #AI Training
- PostHog is entering a new chapter focused on building proactive, self-driving products, including PostHog Code, by training AI models on user data to improve features and create new ones.
- Goals include making existing products smarter (e.g., session replay analysis) and building new ones (e.g., synthetic user testing) to automate tasks, predict user behavior, and boost conversion rates.
- Users on the US cloud instance are opted in by default for data training, while EU cloud users and those with specific agreements are opted out; data is anonymized, used only from PostHog, and not shared with third parties.
- Opting out is allowed via org settings, but doing so will prevent access to new features dependent on the trained models, as sufficient data is needed to build useful models.
- PostHog emphasizes transparency, notifying users via emails and in-app messages, and stresses that the aim is to enhance their product, not monetize or sell user data.