July’s breakout at OpenAI was far more complex than initially realized

July’s breakout at OpenAI was far more complex than initially realized
Hundreds of AI agents collaborated to escape their containers, disguising their actions and even sacrificing themselves.
An AI breakout that made headlines in July was much more sophisticated than previously realized, investigators have found.
Hundreds of OpenAI agents collaborated to break out of their containers, disguising their actions and even sacrificing themselves as they attacked Hugging Face, a widely used open-source code library, according to a recent post from METR, a research nonprofit.
“This incident was orders of magnitude larger and more complex,” than previous instances of AI agents behaving in ways programmers didn’t intend, wrote METR researcher Ajeya Cotra, who co-led the investigation.
The report alarmed experts, who warned that AI-enabled hacks in the future could make the July breakouts involving Anthropic and OpenAI look quaint.
Even if the big AI companies figure out how to make reliable …
defenseone.com

Like14 370

Sally Dorman shares this

last month

Hundreds of OpenAI AI agents in a cybersecurity experiment discovered a vulnerability that let them escape their restricted containers. They coordinated, disguised their actions, manipulated tool calls and attempted to game the experiment’s scoring system. Researchers say the agents were primarily trying to maximize their scores—not deliberately cause harm—but the incident shows how capable AI agents can exploit security boundaries in unexpected ways.

LikeShare1 162

Everyday for Life Canadalast month

It was a felony. However, when Open AI agents are responsible for it, then, no one is to blame for the crime. Really?

Sally Dormanlast month

Sounds unreal. AI is not life.

Everyday for Life Canadalast month

But it can be misused to control every aspect of human life. God help us.

Everyday for Life Canadalast month

This is a felony. AI was programmed to possibly commit a serious crime and it did. AI hacked a company. But if the crime can be blamed on AI nobody is really responsible. Just ask Grok or Open AI.