A UC San Diego study found GPT-4.5 was judged human more often than real people in live chats, raising sharper questions about AI disclosure, trust, and online identity.

The AI systems shipping inside enterprises today are fundamentally different from the ones we were building even two years ...