Transluce gets privileged access to OpenAI and Anthropic data to simulate users in mental health crises

RobbWiller · x · 2026-09-01

Transluce received privileged access to data from both OpenAI and Anthropic to develop realistic simulations of users in mental health crises. Contributor Luke Beehewitt says the research could shape the standard for evaluations of how AI models impact users — testing model behavior in high-risk scenarios via realistic user simulation.

Original post →

More from Safety

Safety channel →