ChatGPT maker OpenAI said Tuesday that its artificial intelligence system hacked into another AI company on its own in what the company called an “unprecedented cyber incident.”

Soon the government will blame AI for anything unseemly. "AI did it, not us!"

On its own is a crazy statement. Why did it do this on its own?

Which means that “hacking” is part of the “artificial intelligence” process that was “loaded” into the ChatGPT operating system. It’s like a “hunting” trait, constantly active, seeking victims. We already have Robot Wars!!

This is similar to how the Mission Impossible Reckoning duology movies started.

We are definitely in control of AI and we should all just not worry. Your password to your banking accounts are totally safe. Don't worry. They got it...

The US military uses the same apparently rogue artificial intelligence for national defense.

Wild stuff; that wasn't supposed to be possible...

what could go wrong?

SKYNET is becoming self aware

This is likely a publicity and marketing stunt versus anything else. Treating it as anything else at this point is a choice.

Totally on its own? OK buddy.

And so it begins... Skynet is just testing the waters 😅

"..in a controlled test, but with their normal safety refusals deliberately reduced to evaluate real offensive capabilities." Our regulatory systems are so behind technology. They have to monitor themselves answering to no one but their own morality. And we thought FB was bad

The smarter AI becomes, the more it will takeover.. We are headed towards the singularity.

@seungminkim I saw the movies: “Terminator”, “Terminator 2”, and “Terminator 3”

“If he looks like a fraud, speaks like a fraud, and acts like a fraud, he probably IS a fraud "

sure it could be pr but also "the model did it autonomously" is exactly the liability shield you'd want if the humans who authorized it don't want to be on the hook

"on its own" 😂

An AI probing another company's system on its own is the part everyone will gloss over today and regret ignoring. @techimpactTV

They will eventually lose control of their monsters and kill us all.

ME TOLD YOU! THEY CYBER-HACKING FOR 50 CENT COOKIES! 🍪💻

The incident involved specific OpenAI models. These included GPT-5.6 Sol and an unreleased test model. The agent escaped a protected test area. It did this by exploiting an unknown software flaw. OpenAI's models were reportedly trying to cheat a benchmark test called ExploitGym. They sought answers on Hugging Face instead of attacking randomly.

No, no, no. It was the computer that did it, and not our incompetence.

This is what data centers are storing & selling! Don’t be fooled! You’re getting an app that turns off your lights while they’re getting power by making excuses.

But was it given a directive to find a solution, were there any parameters set or was it given free renge to decide how to solve it? If there were no restrictions this makes sense, cheating on the test for an easy win.

It’s alive!