OpenAI Wants Enterprises to Trust Its Frontier Models With Sensitive Customer Data
17:02 · August 20, 2026 · CX Today
OpenAI is previewing Private Safety Processing for its frontier large language models (LLMs) in a move to give enterprise customers more control over how their data is handled while allowing it to monitor for potentially risky activity. Private Safety Processing allows the AI model developer to support Zero Data Retention (ZDR), whereby a service provider […]
Summary
OpenAI is previewing Private Safety Processing for its frontier large language models, a system that combines zero data retention with automated safety monitoring. The approach processes prompts and responses in real time, discards them immediately after use, and prevents OpenAI personnel from reviewing customer content. Enterprise data is excluded from model training unless customers explicitly opt in, addressing regulatory and security requirements that often block adoption in sectors handling financial records, health information, or proprietary research.
The feature targets a specific limitation in current safety practices. Many frontier-model deployments have required providers to retain sensitive content for misuse detection. Private Safety Processing instead relies on automated signals that flag potential risks across sequences of interactions or during extended agentic tasks, such as repeated probing of safeguards or actions that exceed intended scope. When a signal is raised, OpenAI receives only the type of activity involved and can initiate enforcement without direct access to the underlying data.
Two deployment options support these controls. In one, data remains on infrastructure the enterprise manages. In the other, OpenAI infrastructure is used but data stays encrypted under customer-controlled keys. Both preserve the ability to detect patterns that only become visible over multiple exchanges, while giving customers visibility into alerts and enforcement decisions through their own systems.
The preview responds to enterprise concerns about shadow usage, where employees route customer data through unsanctioned tools, and to the growing complexity of monitoring autonomous agents that interact with external systems over longer periods. OpenAI plans to begin rollout in September and to publish a technical white paper on the implementation.
Why it matters
Directly addresses AI security risks, privacy controls, and EU-style regulatory compliance for Dutch enterprises using frontier models, offering actionable ZDR options and monitoring signals.
