“Lloyd Blankfein believes a major risk for AI is the inability to test its outputs for correctness, which could necessitate regulatory intervention to slow down deployment.”