OpenAI launches Private Safety Processing to rival Anthropic's data policy
Original: OpenAI seeks to one-up Anthropic with new customer privacy protections
Why This Matters
Enterprise AI adoption increasingly hinges on data privacy; competing retention policies could shift major customer contracts.
OpenAI previewed a new service called Private Safety Processing for select enterprise customers, offering automated cross-session abuse monitoring without retaining any customer data — directly countering Anthropic's 30-day data retention policy for covered models announced in July 2026.
OpenAI has announced a privacy-focused safety feature called Private Safety Processing, currently in preview for select customers. The system conducts automated, long-horizon safety monitoring across multiple user sessions without storing any customer data or involving human reviewers. This approach expands on OpenAI's existing Zero Data Retention (ZDR) policy, which already monitors for abuse on a per-session basis via API agents. Private Safety Processing broadens that scope by analyzing inputs and outputs across multiple sessions to detect coordinated misuse — such as a bad actor spreading malware-engineering requests across separate interactions to avoid detection. If triggered, the system sends a 'narrowly defined signal' to OpenAI indicating a specific type of suspicious activity, after which the company decides on next steps. The announcement is a direct response to Anthropic's July 2026 policy update, which allows the company to retain user session data for 30 days for 'covered models,' including all Mythos-class models like Fable and future models with similar capabilities. That policy, framed as a safety measure, has drawn concern from enterprise customers handling sensitive data. Anthropic otherwise adheres to ZDR for non-covered models. OpenAI's move positions it as a more privacy-preserving alternative for enterprises.