Explore PostHog's plan to train custom AI models using anonymized customer data. See how they're building smarter products and new features.
PostHog Forges Custom AI
Key Takeaways
- 1PostHog will train proprietary AI models on anonymized customer data starting June 29, 2026, to build "self-driving" products like PostHog Code and enhance features like session replay.
- 2PostHog prioritizes data privacy: all training data is anonymized and kept internal, with no third-party sharing. EU users are opted out by default, while US users are opted in but can opt out anytime.
- 3This initiative marks PostHog's shift towards "sovereign AI," emphasizing internal expertise and data control over reliance on external hyperscalers, reflecting a broader industry trend.
- 4Users who opt-in will access advanced, proactive AI features and automated testing, while opting out means foregoing these new intelligent capabilities within the platform.
What This Means For You
1
Review Opt-Out Settings for AI Training
Understand the default opt-in policy for AI model training using your anonymized data. Access your organization settings to opt out if you prefer, noting that some new AI features may not be available if you do.
2
Leverage AI for Proactive Product Insights
Explore how PostHog's new AI capabilities, like enhanced session replay analysis and synthetic user testing, can proactively identify issues and streamline your product development cycle. Prepare to integrate these 'self-driving' features to gain deeper, automated insights into user behavior.
3
Evaluate Sovereign AI Data Practices
Analyze PostHog's approach to internal, anonymized AI model training and its opt-out policies, especially for EU users and those with specific legal agreements. Use this as a benchmark to assess your own organization's data governance strategies for integrating proprietary AI solutions.
4
Consider In-House AI Model Development
Evaluate the strategic benefits of developing proprietary 'sovereign AI' models using internal data, as PostHog is doing. This approach can offer greater control, enhanced data security, and tailored product improvements compared to relying solely on third-party hyperscalers.
FAQFrequently Asked Questions
PostHog announced on May 27, 2026, its plan to train proprietary AI models using anonymized customer data from its platform. This initiative aims to develop smarter products and new features, such as PostHog Code, to enhance the user experience and improve existing functionalities like session replay. The training for these models is scheduled to commence on June 29, 2026.
PostHog is training its own AI models to make existing products smarter, develop new proactive tools, and improve scalability for features like session replay. This approach also enables synthetic user testing to predict potential issues before product release, ultimately aiming for "self-driving" products that offer proactive solutions.
PostHog will manage customer data for AI training by ensuring all data is anonymized and used internally, with no sharing with third-party providers. Users on the EU cloud instance and those with specific legal agreements are opted out by default, while US cloud users are opted in by default but can opt out at any time through their organization settings.
PostHog's shift to "sovereign AI" reflects a growing industry trend where companies prefer customized solutions and greater control over proprietary data, moving away from reliance on large hyperscalers. This approach prioritizes internal expertise, data security, and tailored performance, ensuring that AI models are developed and controlled within the company.
