OpenAI tests 'Private safety processing' that avoids storing customer data
OpenAI is testing "Private Safety Processing" with early customers, a system that can catch misuse patterns but does not store any of your data.
Instead, it just sends over minimal safety signals, so your prompts and responses stay private.
For extra control, businesses can even keep their own data or let OpenAI hold it, with encryption keys they manage.
Anthropic requires 30-day retention
This zero-data-retention setup is mainly for enterprise and API users, not regular ChatGPT plans like Free or Plus.
Aleah Houze, Head of Product Policy at OpenAI, said tracking patterns over time helps spot coordinated threats without sacrificing privacy.
Meanwhile, rival Anthropic has announced a plan to require 30-day data retention on its most capable models for business customers using Fable 5 and Mythos 5, which might not thrill everyone who wants total privacy.
More details on OpenAI's approach are coming in September.