PostHog to train AI models on user data with default opt-in

PostHog has announced a strategic initiative to train artificial intelligence models using data generated on its platform, aiming to enhance features such as session replay analysis and synthetic user testing. The company stated that user data participation is opted in by default for most users, although those on EU cloud instances are opted out by default and must manually opt in. PostHog clarified that the data will be used solely to improve its own products and will not be sold or used to expose models to third parties. Users who opt out will not have access to the new AI-dependent features.

The move represents a significant expansion of PostHog’s existing AI integration efforts, which previously included an AI installation wizard, PostHog AI, and an MCP. The company describes these prior features as wildly popular, though such claims are based on internal statements. The new initiative is closely tied to PostHog Code, a product editor currently in beta, which is designed to create proactive, self-driving products that surface answers and solutions. The company aims to address scalability issues in session replay analysis and automate test and review workloads through synthetic user testing.

PostHog acknowledged that the effectiveness of training models on user data is experimental, with the company admitting that iteration will be required to determine what data is actually useful. It remains unclear how the synthetic user testing feature will perform in identifying user confusion or broken flows before production deployment. The company stated that its goal is to improve PostHog as a product for its customers, rather than to expose or monetise data, emphasising transparency in its approach to this change.

The announcement highlights a tension between the industry-wide push for AI-driven product scalability and growing consumer scrutiny over data privacy. While PostHog asserts that data will be used exclusively to improve its own products and will not be sold, the default opt-in mechanism for most users contrasts with stricter default opt-out protocols for EU cloud instances. This distinction reflects ongoing regulatory complexities and the varying legal frameworks governing data usage across different jurisdictions.

The initiative occurs against a backdrop of broader tech sector developments, including surging demand for AI infrastructure hardware. Micron Technology recently reported a 196 per cent year-on-year revenue surge driven by AI data centre demand, underscoring the high stakes of balancing innovation with user trust. PostHog’s decision to be upfront about data usage, rather than burying it in terms and conditions updates, positions the company as seeking to maintain transparency in an increasingly sensitive area of technology development.