NEWS
 OpenAI launches Control Room with a powerful 35-second kill switch for rogue agentsOpenAI has launched Control Room to monitor and halt risky AI workloads in its research infrastructure via a centralized system. The system strengthens security measures by preventing the model from generating new outputs within 35 seconds in the event of a security concern. This is a news summary compiled for the AiPulse newsletter — not the original article. The original article is at interestingengineering.com. Read the original article AI Critique → English (automatic translation) |