OPENAI RESUMES FRONTIER AI TRAINING AFTER SECURITY OVERHAUL , DISCLOSES FLAW IN WIDELY USED SOFTWARE

What happened: OpenAI resumed training of its most advanced AI models after tightening security and barring internet access, and disclosed a flaw in Brotli, a widely used data compression tool, found while testing its defenses. Key numbers: Training resumed October 1. The Brotli root cause was fixed in February. Why it matters: The disclosure highlights concerns over AI agents identifying and exploiting software flaws, including in the guardrails meant to contain them. METR and Redwood Research reported separate vulnerabilities in OpenAI's internal system. Coverage: 5 outlets in 24h

OpenAI resumed training of its most advanced AI models after tightening security and barring internet access, and disclosed a flaw in Brotli, a widely used data compression tool, found while testing its defenses.

The disclosure highlights concerns over AI agents identifying and exploiting software flaws, including in the guardrails meant to contain them. METR and Redwood Research reported separate vulnerabilities in OpenAI's internal system.