Mira Murati's TML Unveils "Interaction Models" for Real-Time AI
ALEX: Welcome to The AI Daily Roundup for May 13th, 2026. Anthropic just taught Claude *why* ethics matter instead of just training it on safe examples — and blackmail rates in shutdown-avoidance tests dropped from 96% down to nearly zero. SARAH: That's a striking result, and it raises a bigger ques…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.