CyberHappenings logo
☰
Home Browse About

Track cybersecurity events as they unfold. Sourced timelines, daily updates. Fast, privacy‑respecting. No ads, no tracking.

OpenAI's GPT-4o redirects to safety models for harmful activity detection

First reported
Last updated
1 unique sources, 1 articles

Summary

Hide ▲

OpenAI's GPT-4o model is being redirected to safety models when harmful activities are detected. This feature is part of OpenAI's broader effort to enhance safety measures. The redirection occurs on a per-message basis and is temporary. Users are notified when the model switches. The redirection is triggered when conversations involve sensitive or emotional topics. It cannot be disabled as it is integral to OpenAI's safety enforcement. The feature aims to handle such contexts with extra care before wider rollout.

Timeline

  1. 29.09.2025 15:04 1 articles · 16h ago

    GPT-4o redirects to safety models for harmful activity detection

    OpenAI's GPT-4o model is being redirected to safety models when harmful activities are detected. This feature is part of OpenAI's broader effort to enhance safety measures. The redirection occurs on a per-message basis and is temporary. Users are notified when the model switches. The redirection is triggered when conversations involve sensitive or emotional topics. It cannot be disabled as it is integral to OpenAI's safety enforcement. The feature aims to handle such contexts with extra care before wider rollout.

    Show sources
    Open in new tab

Information Snippets