Tech & SciencePublished

OpenAI restarts frontier AI training with new safety limits

OpenAI has resumed training its most advanced AI models, limiting their internet access and adding tighter security controls, according to CNA, while Rappler reports the restart follows a security overhaul that included disclosing a flaw in a common software component and raises worries that powerful AI could exploit such vulnerabilities.

Summarised with AI from 2 sourcesRead our AI policy

Why it matters

It shows heightened safety steps as AI capabilities expand.

What changed

  1. First published from CNA and Rappler