
OpenAI Adds Security Checks That Cost 20% Extra Computing Power
OpenAI announced August 18, 2026 new security measures for its AI systems: isolating networks so a single break-in cannot spread, and monitoring AI activity with alerts targeted within 30 minutes. The monitoring requires about 20% extra computing power. VP of research Amelia Glaese said safeguards will tighten as AI systems become more capable.
Published