PostHog Plans to Train AI Models on Customer Data to Power New Product Intelligence Features
Key Takeaways
- ▸PostHog Code, a new product editor launched in beta, will be powered by AI models trained on anonymized customer data
- ▸The company will use trained models to improve session replay analysis at scale, enable synthetic user testing, and provide proactive product recommendations
- ▸Data governance approach: US users opt-in by default, EU users opt-out by default, with universal opt-out available anytime; all data is anonymized before training
Summary
PostHog announced plans to train its own AI models on customer data to power the next generation of AI-driven features, including the newly launched PostHog Code—a product editor designed to help teams build better products faster. The company will use anonymized data from existing customer instances to improve session replay analysis, enable synthetic user testing to catch bugs before production, and predict user behavior changes that could boost conversion rates. PostHog is adopting a transparent approach to data governance: US cloud users are opted in by default, EU cloud users are opted out by default (complying with stricter regulations), and all users can opt out anytime via their organization settings. Notably, PostHog will conduct all model training in-house and will not sell or send customer data to third-party AI providers—a deliberate choice to retain control and trust. Model training will begin June 29, 2026, giving customers a full month to review their settings and make decisions.
- PostHog will train models in-house and will not monetize customer data through third-party licensing or provider sales
- Training begins June 29 with advance notification via email and in-app messaging to all affected users
Editorial Opinion
PostHog's decision to train models on customer data reflects the industry's growing comfort with AI-powered product improvement, but their transparency-first approach is notably different from competitors who quietly slip data clauses into terms of service. The opt-in-by-default policy for US users diverges from EU privacy norms, signaling PostHog's prioritization of product velocity over precaution—a bet that customers trust them enough to accept the trade-off. The commitment to in-house training rather than outsourcing to third-party LLM providers shows PostHog believes competitive advantage lies in understanding user behavior, not in building AI models themselves.
