February 19, 2026 | wdelong | 1 min read
A study led by MIT researchers found that agentic AI developers seldom publish detailed information about how these tools were tested for safety.