Nobody Knows How to Safety-Test AI

Governments and companies hope safety-testing can reduce dangers from AI systems. But the tests are far from ready.

Full story: https://time.com/6958868/artificial-intelligence-safety-evaluations-risks/