Introducing Lockdown Mode and Elevated Risk labels in ChatGPT

Source: OpenAI Date: 2026-02-13 URL: https://openai.com/index/introducing-lockdown-mode-and-elevated-risk-labels-in-chatgpt

Summary

Title-only: ChatGPT adds two safety features: “Lockdown Mode” (likely a high-security mode that restricts AI capabilities for users who may be targeted by sophisticated attacks — analogous to iOS Lockdown Mode for journalists and activists) and “Elevated Risk labels” (warnings attached to conversations or queries that are flagged as high-risk based on content). February 2026 places this in the period of OpenAI’s concerted security hardening push.

Implications

The user safety feature thread. Lockdown Mode for ChatGPT is a significant product decision: it implies OpenAI has identified user categories (activists, politicians, researchers, journalists) who face elevated risk from AI-powered attacks and need a restricted-capability, high-security mode. This is OpenAI acknowledging that their product can be weaponized against their own users — and building defensive features accordingly.

Elevated Risk labeling. The “Elevated Risk labels” feature is particularly interesting as a UX decision: it surfaces risk assessments within the conversation rather than silently refusing or complying. This creates user awareness of AI risk in real-time, which is both protective (users can recalibrate) and educational (users learn what constitutes elevated risk). The transparency approach contrasts with pure content blocking, which can feel arbitrary.