OpenAI Wraps Business AI in Privacy Controls

The first consequence is contractual, not technical: the moment a vendor offers a privacy tier, every deployment without it becomes a decision somebody signed.

OpenAI's Private Intelligence, reported in this magazine's October 31 Night edition, packages two controls enterprises have been asking for since the first chatbot pilot: zero data retention — for eligible API workflows, prompts and model responses are not kept after processing — and private inference, so the computation itself runs under the customer's privacy terms. Zero data retention is available now; private inference is in preview, which means the controls exist and the paperwork is catching up.

For a manager, the plot is familiar from every cloud negotiation of the last decade. The feature list is the easy part; the contract is where privacy actually lives. Which data classes are covered, whether logs and abuse-monitoring copies count as retention, which subprocessors touch the inference path — those answers decide whether this is a control or a brochure.

The timing is the story. In the same week, an AI-agent campaign hit 200 organizations at machine speed. Boards will ask two questions — are we protected from AI, and are we protected in our use of AI — and this launch is built to answer the second one in the room where budgets are approved.