OpenAI Rolls Out Privacy-Focused Safety Tool in Apparent Jab at Anthropic
OpenAI has unveiled a new automated monitoring system that flags potential AI misuse without storing customer data, contrasting with Anthropic's recent data-retention policy.
OpenAI has announced a preview of a new service called Private Safety Processing, currently being tested with select customers. The system automatically monitors for potential misuse of its AI models while retaining none of the customer's data, positioning it against a policy recently adopted by rival Anthropic.
In July, Anthropic introduced a policy allowing it to retain user conversations for up to 30 days for so-called "covered models," including all Mythos-class models. The policy was intended to support safety analysis but has unsettled some enterprise customers who handle sensitive data and don't want it stored or reviewed by the AI lab.
How the new system works
Like most AI companies, OpenAI generally follows a Zero Data Retention (ZDR) approach, using automated agents to check for abuse within individual sessions without keeping user data. The new Private Safety Processing technology extends this by analyzing multiple conversations together rather than just one session, allowing it to catch patterns where a bad actor deliberately spreads requests across sessions to avoid detection.
When the system detects suspicious activity, it sends OpenAI a narrowly defined signal about a specific type of behavior, without any human review of the conversations themselves. Based on that signal, OpenAI decides whether enforcement action is needed — for example, contacting the customer, who may then choose to share additional data at their own discretion.
Anthropic, for its part, says human review of data does occur in its system, but only through a controlled access path involving a small set of approved reviewers, with every review session logged in a tamper-proof record.
The announcement comes amid intensifying rivalry between the two companies. A recent report indicated OpenAI's second-quarter growth lagged behind Anthropic's, whose annualized revenue run rate has reportedly reached around $65 billion. Both companies are also reportedly working toward eventual public listings.
