ChatGPT Officially Rolls Out Locked Chats: Biometric Authentication and Complete Memory Isolation Across All Tiers

Background and the Problem

As frontier AI assistants embed deeply into enterprise workflows, features like persistent cross-session memory introduce significant privacy hazards. Users routinely interleave trivial queries with highly confidential activities—such as proprietary software vulnerability triage, corporate restructuring discussions, and sensitive health evaluations. Previously, without turning off global memory entirely, facts extracted from confidential sessions were automatically indexed into the user's permanent memory bank, periodically bleeding into unrelated subsequent prompts. Furthermore, physical device sharing exposed unlocked conversation histories. Addressing these data governance vulnerabilities, OpenAI officially launched Locked Chats (@ChatGPT, Status #2112597766333673624) on October 21, 2026.

Architecture and How It Works

1. Native Biometric Gating:

• Integrates directly with device-level secure hardware enclaves (iOS Face ID, Android Biometrics, Windows Hello, macOS Touch ID);
• Locked threads are immediately sealed behind a secondary authentication layer, requiring biometric verification to view or search;

1. Absolute Memory Isolation:

• Implements strict architectural write-fences: token streams, entity extractions, and behavioral observations generated within locked sessions are blocked from persisting into long-term user memory;
• Ensures zero cross-context contamination between confidential deliberations and routine everyday prompting;

1. Stealth UI & Background Protection:

• Automatically hides locked threads from the sidebar hierarchy and suppresses notifications, blurring window contents immediately upon app switching.

Benchmarks and Measured Results

Disclosed in OpenAI's privacy architecture briefings and community testing:

• Memory Bleed Zero-Leakage Verification:
• Across adversarial prompt-injection probes probing cross-session fact recall, information contained within locked chats recorded a 0.00% memory leakage rate;
• Universal Availability:
• Accessible without tier restrictions across Free, Plus, Team, Enterprise, and Edu user bases;
• Local authentication unlock latency clocks under 150 ms, preserving fluent productivity.

Getting Started for Developers

• Activation: In the ChatGPT mobile app or desktop web interface, right-click or long-press any conversation and select 'Lock Chat';
• Official Announcement: View verified feature walkthrough via @ChatGPT status https://x.com/ChatGPT/status/2112597766333673624;
• Operational Recommendation: Engineering and corporate teams utilizing ChatGPT for proprietary code refactoring or security assessments should mandate Locked Chats as standard operating procedure (SOP) to ensure uncompromised IP protection.