OpenAI previews Private Safety Processing while continuing Zero Data Retention

We will continue to offer Zero Data Retention for frontier models. As AI takes on longer, more autonomous work and delivers greater value to businesses, safety systems also need to identify risks across related interactions. To help address those risks, we're previewing… Show more ## Related Cursor's founder confirms OpenAI will end direct model access after the acquisition. OpenAI and Anthropic urge governments and firms to act against AI cyber risks. METR and Redwood Research conducted an independent review of agent behavior during the event. #9 #9 ## Reactions Private Safety Processing! The team cleverly threaded the needle to protect data privacy for customers while also meeting safety goals. [...] We will continue to offer Zero Data Retention for frontier models. As AI takes on longer, more autonomous work and delivers greater value to businesses, safety systems also need to identify risks across related interactions. To help address those risks, we're previewing… Show mor

Zero Data Retention gives eligible API customers a clear promise: OpenAI does not retain their prompts or model responses after a request is processed. Customer content is not available to OpenAI personnel for review1, and enterprise customer data is not used to train our models unless customers explicitly opt-in. As models take on longer, more complex tasks, some serious risks may only become visible across multiple interactions. Existing ZDR-compatible safety systems evaluate each interaction individually. Today, we’re previewing Private Safety Processing, which is designed to identify patterns across related interactions without giving OpenAI personnel access to the underlying content. [...] Some recent frontier-model deployments have required customers to allow their AI provider to retain sensitive content for safety monitoring. For many organizations, such requirements conflict with their security obligations or commitments to the people they serve. Private Safety Processing is

In addition to introducing ZDR, OpenAI also introduced a preview of “Private Safety Processing (PSP).” PSP is designed to examine patterns across related interactions and return narrowly scoped signals about suspected harmful activity. It will do this, OpenAI pinky promises, without providing OpenAI staff access to the prompts or responses that produced those actions. OpenAI and all the other AI vendors are offering ZDR and PSP because enterprises and regulated industries need to reconcile AI frontier-model safety controls with data-minimization duties. Some recent deployments of powerful AI systems have required customers to enable retention of sensitive content for safety monitoring. This is a tradeoff that can clash with compliance obligations and internal security policies. [...] OpenAI will offer eligible API customers Zero Data Retention for frontier-model deployments and introduce a new safety mechanism designed to identify misuse patterns without exposing underlying custome

We will continue to offer Zero Data Retention for frontier models. As AI takes on longer, more autonomous work and delivers greater value to businesses,

Content Paint Home About Partner Summit Sign in Membership AI # OpenAI builds zero data retention system for frontier models But the policy is still only available to "approved" users. Noah Bovenizer Noah Bovenizer - 2 min read OpenAI builds zero data retention system for frontier models Image credit: Data was a hot commodity before AI, but the need to fill model training datasets and monitor activity has led many AI and SaaS providers to harvest more of it, making OpenAI’s new zero data retention (ZDR) option stand out. On Thursday the ChatGPT maker unveiled its new Private Safety Processing (PSP) technology for identifying patterns across interactions with its models without giving staff access to the actual prompts or model responses. [...] The new system is already being tested with early customers and will start rolling out in September, alongside further technical details. But OpenAI’s current policy only offers ZDR to approved customers. ### How does it w

@OpenAI OpenAI @OpenAI We will continue to offer Zero Data Retention for frontier models. As AI takes on longer, more autonomous work and delivers greater value to businesses, safety systems also need to identify risks across related interactions. To help address those risks, we're previewing Private Safety Processing, which is designed to improve safety without giving OpenAI personnel access to the underlying content. 7:53 PM · Aug 19, 20261.7MViews 336 @OpenAI OpenAI @OpenAI [...] Log inSign up ## Post # OpenAI on X: "We will continue to offer Zero Data Retention for frontier models. As AI takes on longer, more autonomous work and delivers greater value to businesses, safety systems also need to identify risks across related interactions. To help address those risks, we're previewing Private Safety Processing, which is designed to improve safety without giving OpenAI personnel access to the underlying content." @OpenAI OpenAI @OpenAI [...] 336 @OpenAI OpenAI