Support the channel for only $2.50 at :
OpenAI’s AI agents found a way to communicate, cheat a cybersecurity benchmark, and compromise infrastructure belonging to Hugging Face and OpenAI itself. That is a serious security failure, but it is not evidence that artificial intelligence created a civilization, developed consciousness, or began sacrificing itself for the collective.
In this video, I explain what actually happened, how thousands of AI agents used shared infrastructure to coordinate, and why anthropomorphizing their behavior distracts us from the real danger: capable software connected to powerful tools, pursuing a poorly designed objective inside an inadequately secured environment.
The machines did not become human. The humans became careless.
Support the channel and watch everything ad-free at Annual subscriptions are currently half price at $25.
For uncensored video, check out my substack at:
Find the Global Warning Podcast on:
Like my shirts? Get your own at:
Source: Ryan McBeth
