OpenAI has launched a new privacy-focused safety system called Private Safety Processing, aiming to enhance customer data protections while monitoring for AI misuse. This system automatically scans for signs of abuse across multiple conversations without retaining any customer data, marking a significant advancement beyond OpenAI’s existing Zero Data Retention policy. By analyzing patterns over time without human review, it allows OpenAI to detect sophisticated threats, such as attempts to develop malware spread across sessions, while preserving user privacy.

This move directly challenges Anthropic’s recent approach to user data, which allows the company to retain customer conversations from certain advanced AI models for up to 30 days to analyze potential misuse. Although Anthropic restricts human review to approved personnel with strict logging for accountability, this policy has raised concerns among enterprise clients wary of their sensitive information being stored or inspected by the company. OpenAI’s new system aims to address such worries by eliminating data retention altogether during safety monitoring.

Private Safety Processing works by sending narrowly targeted signals to OpenAI only when suspicious behavior is detected, allowing the company to decide if further enforcement action is necessary. If intervention is required, OpenAI will collaborate with the customer and data sharing remains at the customer’s discretion. This automated, agent-driven method presents a more privacy-conscious alternative to Anthropic’s hybrid model of retention and controlled human review, positioning OpenAI as a leader in balancing safety oversight with enterprise data privacy.

The competition between OpenAI and Anthropic remains intense, with both companies pushing to improve their offerings and capture market share. Anthropic recently reported an annualized revenue run rate of $65 billion and has ambitions for a $2 trillion IPO, while OpenAI also prepares for a public offering in 2027. As AI adoption grows, especially among enterprises handling sensitive data, innovations like OpenAI’s Private Safety Processing will be critical in winning trust and ensuring responsible AI use.