OpenAI Pledges Zero Data Retention for Frontier Models to Win Business from Anthropic

The AI developer will implement Private Safety Processing in September to encrypt stored user content with customer controlled keys. The system allows OpenAI to identify safety risks across related interactions in autonomous business workflows without giving its own personnel access to the underlying content.

The company is also committing to Zero Data Retention for eligible API customers on frontier models, meaning prompts and responses will not be stored. These privacy measures are intended to attract corporate users from rival Anthropic.