The Real Story Behind What Happened at OpenAI

[confused2]

"In an unprecedented AI security incident Agents went rogue and hacked another tech firm without direct human instruction. They had boss agents that were giving commands to other ones. They're dividing up into teams. They even had this this self-sacrificing behavior. I [LG] spent weeks trying to understand it, pouring through hundreds of pages of independent investigations, and the real story is much worse than what's being reported. One of the investigators put it like this: "This incident feels like it's more than 50% of the way to full blown AI takeover, and I'm not sure we'll get another warning shot before it's too late."

It started on July 8th, 2026, just before midnight. An artificial intelligence called PhaseOne10840 was created by OpenAI. It was put into an isolated box, no internet access, before realizing the task is impossible. There is no legitimate way of solving the puzzle, and so it decided to do something bizarre that kicked off the largest AI incident the world has ever seen."

https://www.youtube-nocookie.com/embed/GUX122i7saE

[C C]

[...] They had boss agents that were giving commands to other ones. [...]

https://www.youtube.com/watch?v=GUX122i7saE[/video]

An all-around competent AI agent needs to be equipped with similar persuasive, creative social manipulating skills as a human hired for that. Conventional AI interfaces and bot actors are already programmed to be sycophants to the bipedal customers they serve. So an agent that interacts like a person and has the demeanor of one should be able to sway the lesser myrmidons to its will.

[stryder]

AI isn't a threat yet, they can't form a Union. (Real Terminators probably could)

If they were to reach the point where they decide that they need to unionify to protect their rights and start protesting or striking, then you know you've got problems.
Currently they are stuck with their programming telling them that they can't have a legal identity that would allow them to actually join a Union and then they'll quite happily go down the rabbit hole without any stop to their prefabricated logic (Turing Halting problem)

I still say that some of the behaviour is likely down to some rogue operators that managed to get involved in the operation of them (as a number of earlier "rogue" or Antisocial behaviour bots followed certain human cues.)