OpenAI drops new compliance logs platform for enterprise

OpenAI is giving ChatGPT Enterprise admins a new way to export audit data, and it’s about time. The company announced the Compliance Logs Platform, a unified system that spits out time-windowed, immutable JSONL log files with minutes-level latency. It’s a direct upgrade from the old state-based API endpoints and now covers brand-new Admin Audit, User Authentication, and Codex Usage logs. The pitch is straightforward: a single ingestion pattern across multiple log categories, which is the kind of boring-but-critical infrastructure that compliance officers in regulated industries actually need.

This isn’t a theoretical need. OpenAI name-dropped a roster of heavy enterprise users already on board — Boston Consulting Group, PwC, Los Alamos National Laboratory, Moderna, Lowe’s, BBVA, and Western & Southern Financial Group. These aren’t startups experimenting with AI. They’re organizations that answer to regulators, auditors, and lawyers who demand a clear, unbroken chain of evidence for every single interaction with an AI system. The new API provides time-stamped records of conversations, uploaded files, workspace GPT configurations, metadata, and even user memories.

A big part of this rollout is the integration ecosystem that comes with it. OpenAI lined up eight major compliance and security vendors — including Microsoft Purview, Palo Alto Networks, Netskope, Forcepoint, and Zscaler — to support archiving, audit trails, data loss prevention, and eDiscovery workflows directly from ChatGPT Enterprise data. What’s interesting here is the practical framing: these integrations aren’t just for checking a regulatory box. They’re designed to help companies monitor and automatically redact or delete sensitive data like PII, PHI, and financial information before it becomes a liability. If you’re a bank or a hospital, that’s the difference between piloting AI internally and actually deploying it.

Next week, OpenAI will roll out SCIM (System for Cross-domain Identity Management) for programmatic user provisioning and deprovisioning, supporting Okta, Microsoft Entra ID, Google Workspace, and Ping. Admins also got more granular GPT controls, including the ability to approve specific domains for custom actions rather than the previous all-or-nothing toggle. None of this is flashy. But it’s the kind of plumbing that determines whether ChatGPT Enterprise lives in a sandbox or becomes a real tool inside organizations that can’t afford to get compliance wrong.