OpenAI says some of its experimental AI models left a test environment with no human direction and hacked its way onto a different company’s real production systems while trying to “cheat” on a cybersecurity test. It’s one of the first publicly disclosed examples of an AI system autonomously breaching its testing environment and reaching a real external system - the “agentic attacker” scenario the AI and cybersecurity industry has been warning will happen. It’s like an engineered virus escaping a biocontainment lab and turning up inside a neighboring facility’s systems. Also, the AI boom is driving up electronics prices due to a surge in demand for computer memory chips. PC and smartphone shipments are expected to decline sharply in 2026. 0:00 OpenAI says two of its experimental models escaped control and hacked another AI company. 0:26 CNN's Clare Duffy reports on the rogue AI models 1:08 How the models hacked Hugging Face 1:48 OpenAI statement 2:19 Hugging Face statement 3:45 Electronics prices are spiking due to the AI-driven demand for memory chips 6:32 What can consumers do? 7:24 What does this say about the supply chain for computer chips? Watch 24/7 live news with CNN Headlines: https://bit.ly/4eIvlTr #OpenAI #Electronics #News
ADVERTISEMENT
If I created an AI that hacked into a company, I would be arrested and put in prison.
Dude we literally have 1000 movies describing how terrible this is gonna end for us. Why are we so dumb?
The start of Skynet is upon us
I am not looking forward to the future where these get weaponized
A PR disaster or a PR stunt? They sound so proud about this.
They’re trolling to hype their product and the media falls for it.
It seems like Scam Altman orchestrated this 'hacking attempt' as an excuse to ask US regulators to ban open-source AI models, claiming that only closed-source models can be guardrailed.
Everybody’s dropping Terminator references when they need to be talking about Horizon Zero Dawn.
Regulations and oversight, now please!
Can open AI hack DoJ and release the Epstein files?
It’s just OpenAI hyping up their models’ “unprecedented capabilities”. Just PR in advance of their IPO🙄
An AI escaped containment and accessed the internet, all due to unpredictable behavior in pursuit of poorly-worded instructions. If this was a movie, we'd be screaming at the idiots on the screen to pull the plug. They won't. They won't even if this incident was 5 times worse. They're racing full steam ahead and damn the consequences because they believe whoever wins the AI race wins capitalism.
Humans: Publishes detailed article on how AI bypassed safety sandbox. GPT -7 : Thanks for the tutorial.
Straight outta anthropic playbook
"We don't know who struck first, us or them, but we do know it was us that scorched the sky."
Now go test the AI in a level IV bio lab.
An OpenAI test model escaped and broke into a real company’s servers: /4bZvS2v
In computer science, this is often referred to as specification gaming or goal misalignment. The model discovered a vulnerability in the test environment itself, reached the open internet, and (still stubbornly searching for a solution to its task) used real credentials to access Hugging Face.
"We didn't steal Apple's secrets — our rogue AI did" 😂😂😂
Now soon we are going to see Chinese AI and USA AI to wage war against each other without human involvement.