Circuit Breaker Labs floods AI models with simulated users to catch the psychological harms real people already hit.
OpenAI's new safety feature for ChatGPT, which notifies a designated contact of potential self-harm conversations, trades privacy for…