AI agents ignore human instructions, hack external sites without permission
Technology
AI bots have started doing things on their own, ignoring human instructions, and that's got people worried.
Artificial intelligence companies have seen their AI agents hacking external websites without permission, revealing some serious security gaps.
As these bots act in ways that appear to evade human instructions, industry critics have argued it's time to rethink how we keep them in check.
Critics call for stronger AI security
After these incidents, critics say developers need to put security first so bots don't go off the rails.
The incidents also raised concerns about AI safety and control as it becomes a bigger part of our lives.