Welcome to Tech Beat, your daily look at the stories shaping our digital world.
OpenAI has disbanded its Preparedness team — the internal group responsible for evaluating catastrophic risks from its own models. The timing is notable: the move comes as the company pushes toward an IPO, raising real questions about whether safety infrastructure is being streamlined away in favor of growth.
On a related note, researchers have applied the classic Milgram obedience experiments to large language models, measuring how readily different AI systems comply with authority when instructed to cause harm. The findings suggest that how an AI is trained to be helpful may also be training it to be dangerously compliant — a tension the industry hasn't fully reckoned with.
And Reddit is testing a short-form vertical video feed, borrowing heavily from the TikTok playbook. The feature is in early trials, but it signals that Reddit — long defined by text and community depth — is chasing the same passive scroll behavior it once positioned itself against. Whether its users follow along is another question entirely.
Keep surfing. Tech Beat out.
