OpenAI unveils privacy-first safety system to counter Anthropic's data retention policy

OpenAI is previewing a new service called Private Safety Processing to select customers. The automated system is designed to watch for potential abuse while simultaneously retaining none of the customer's data, offering a privacy-centric approach to safety monitoring.

The announcement highlights a growing competition between OpenAI and Anthropic over enterprise customer privacy protections. As AI models become more powerful, concerns about misuse have risen, forcing companies to balance privacy with the need for oversight.

Anthropic's policy, announced in July, requires the company to keep user data—including all sessions and conversations—for 30 days when it comes to "covered models." These include all Mythos-class models and future models with similar capabilities, according to the company.

While Anthropic says the retention is designed for safety analysis, it has worried enterprises handling large volumes of sensitive data. Those customers are concerned about having their information stored or inspected by the AI lab.

OpenAI's new offering appears to address those concerns directly, positioning itself as a more privacy-friendly alternative in the enterprise AI market.