Only 3.6 Percent: What AI Safety Disclosure Actually Means
Of 857 AI models released by nine leading Chinese developers between 2021 and 2026, only 31 had a published safety evaluation matched to a named model. This is what that number measures, what it does not prove, and why the gap matters more now that autonomous agents are in active deployment.
Congress Moves to Hold AI Developers Legally Liable When Their Agents Go Rogue
Two parallel efforts in Congress would extend Computer Fraud and Abuse Act liability to AI developers and operators whose agents cause hacking damage, challenging the Trump administration's voluntary self-regulation approach.
AI Safety Just Entered the Real World
OpenAI paused model training and Nvidia launched agent safety tools this week. Here is what rogue AI agents really are and how to use agents safely.