OpenAI rolls out Privacy Filter to scrub sensitive data from model interactions

OpenAI announced a Privacy Filter feature designed to detect and redact personally identifiable information and other sensitive data before it reaches model training pipelines or persisted logs. The filter operates as a preprocessing layer, intercepting prompts and outputs to flag categories like names, addresses, financial identifiers, and health-related disclosures.

The move responds to growing enterprise concern about inadvertent PII leakage when employees paste internal data into ChatGPT and API calls. By shifting redaction upstream of storage, OpenAI aims to reduce the blast radius of any future log breach and simplify compliance posture for customers in regulated sectors.

The available source content is minimal — only the headline was retrievable — so specifics on detection accuracy, opt-out behavior, and whether the filter applies to consumer ChatGPT or only enterprise tiers are not confirmed here. Treat the framing as directional pending the full announcement.