2026-04-27 19:15:00+08
In a major update to its safety suite, OpenAI has launched "Privacy Filter 2.0," which now supports an ultra-long context window of 128,000 tokens. This allows enterprises to process massive documents—such as entire legal contracts or medical histories—through the privacy layer in a single pass, ensuring that every piece of sensitive information is identified and anonymized before further analysis.
Version 2.0 introduces eight new types of "Privacy Recognition," including advanced biometric data, cryptographic keys, and subtle "Contextual Identifiers" that could be used to de-anonymize a user. The filter is powered by a new, high-speed model that adds minimal latency to the AI pipeline, making it suitable for real-time applications like customer support chatbots.
OpenAI’s CTO stated that the 128K context window was the most requested feature from enterprise partners. "We are moving toward a future where large-scale data processing doesn’t have to mean large-scale privacy risk," they said. The update is available immediately for all API users and is integrated into the latest ChatGPT Enterprise tier.