Introducing OpenAI Privacy Filter

Source: OpenAI Date: 2026-04-22 URL: https://openai.com/index/introducing-openai-privacy-filter

Summary

OpenAI’s April 2026 launch of the OpenAI Privacy Filter — a feature or setting allowing users to prevent their conversations from being used for training, limit data retention, or apply stricter privacy controls to their ChatGPT interactions. Published the same day as the Workspace Agents launch and the WebSockets Responses API update, the Privacy Filter addressed the tension the Personalization guide had highlighted: as ChatGPT stored more user context and memory, privacy controls became increasingly important for user trust.

Implications

Privacy controls as a competitive necessity. Anthropic and other competitors had offered training opt-out and privacy controls as defaults or easy-to-enable settings, creating expectations among privacy-conscious users that OpenAI had to meet. The Privacy Filter also addressed enterprise procurement requirements — organizations in regulated industries needed contractual and technical controls over how employee-generated data was handled.

Thread: ChatGPT privacy and user data. Sits alongside the “Our Approach to Data and AI” post (May 2024), the NYT privacy counter-claim (November 2025), and the data residency expansion (November 2025) as OpenAI’s ongoing effort to demonstrate responsible data handling as the product’s data collection scope expanded with memory and personalization features.

Watch: What the Privacy Filter actually prevented — training data use, conversation logging, memory retention — and whether the controls were granular enough to satisfy enterprise security requirements and GDPR deletion obligations.