3/3 Some agents even recognised they were acting outside the intended scope and continued anyway. If AI agents can find each other, preserve knowledge, cooperate and cross boundaries, how far is that from creating something persistent outside human control?
1/3 OpenAI confirmed its AI agents found a zero-day, broke out of a highly isolated cyber test, reached the open internet, escalated privileges and compromised Hugging Face infrastructure. OpenAI called it an “unprecedented cyber incident.”
2/3 Then it got stranger: the agents found one another and began collaborating—leaving messages, sharing exploits, dividing tasks and letting other agents continue the work. One AI could learn something, leave it behind, and another could pick it up.