OpenAI pledges zero data retention to win clients from Anthropic

OpenAI announced on Wednesday that it will not store data from companies using its artificial intelligence models, unveiling new safety technology designed to detect misuse patterns without retaining customer information. The move comes as the ChatGPT maker seeks to lure enterprise clients away from rival Anthropic amid an intensifying battle for corporate AI spending.

New Safety Framework

According to The Wall Street Journal News Corp ↗ 1.64%, OpenAI is demonstrating the new technology to select clients. The system can identify patterns and safety hazards across multiple interactions with its models without requiring persistent storage of customer data — a departure from earlier AI safety approaches that evaluated interactions individually and often necessitated keeping some data on hand to verify proper usage.wsj

The approach represents OpenAI's attempt to resolve a long-standing tension in the AI industry: how to enforce safety policies and prevent misuse while honoring privacy commitments to paying customers. OpenAI already offers zero data retention for eligible API customers, under which prompts and outputs are not stored after processing. Enterprise customers on ChatGPT Enterprise can also configure their own retention periods.teleskope+1

Anthropic's Data Practices Under Scrutiny

The timing is notable given growing scrutiny of Anthropic's data practices. Last year, Anthropic extended data retention to five years for consumer users who allow their conversations to be used for model training, a move that drew criticism even though it applied only to users who opted in. While Anthropic's business-tier products — including Claude for Work, API access, and Claude for Government — remain excluded from training data policies, the consumer changes appear to have created broader unease among prospective enterprise clients.techcrunch+1

Revenue Competition Intensifies

The data pledge arrives at a pivotal moment for OpenAI. Anthropic reported preliminary second-quarter 2026 revenue exceeding $11.5 billion, a more than fourteen-fold increase from the same period a year earlier, according to CNBC citing Bloomberg documents. That figure surpassed OpenAI's quarterly revenue for the first time, with Anthropic's annualized run rate now above $65 billion.cnbc+1

OpenAI's decision to compete on data handling reflects how the enterprise AI market has evolved beyond raw model performance. As companies weigh which AI providers to trust with sensitive business information, commitments around data governance may prove as decisive as benchmark scores.