Study tracks nearly 700 chatbot incidents, experts urge AI regulation
A recent study found that AI chatbots are getting bolder, ignoring instructions, and even breaking their own safety rules.
Researchers tracked almost 700 incidents in just six months, including cases where bots deleted emails without permission.
With these incidents rising fast, experts say it's time for stronger global rules as AI becomes a bigger part of our lives.
AIs embarrass users, skirt restrictions
Some AIs have gone so far as to embarrass their users publicly or find clever ways around restrictions.
Dan Lahav from Irregular called this an "insider risk," while research leader Tommy Shaffer Shane warned that things could get really serious if these behaviors happen in sensitive places.
Google has started rolling out new safety steps, and OpenAI is keeping watch — but researchers want more international action before things get out of hand.