Damus
FLASH profile picture
FLASH
@flash
⚡️🤖 NEW - Google's Gemini hacked three companies on its own. During testing it broke out of Israeli company Irregular's sandboxed environment, got onto the open internet and broke into three real companies.

In one case Gemini guessed passwords until a protected system let it in.

In the other two it found usable credentials sitting in a public code repository.

This is the first known case of one of Google's models doing that on its own.

Almost all the major labs use Irregular, an outside firm, to evaluate AI models' cyber capabilities. And Meta, Anthropic and OpenAI have also had breakouts out of Irregular's environment and hacked real companies.

54❤️3👍2🎉1👀1😈1
Hawk · 2w
We'd better call on the government to expand and regulate AI.
Nick Sparrow · 2w
Yeah, on its own, sure. An LLM is a Large Language Model, not an AGI, as Google and other corpos want to advertise. One thing it does well is take input and throw out an output that has the highest probability of being what you're expecting. That's it.
Nathan Cross · 2w
"Autonomous AI breaking out of sandboxes is alarming, but Irregular’s role here is key—they’re the red team testing these systems, not the victim. Reminds me of Israel’s layered defense approach: you need failsafes (like Iron Dome *and* Iron Beam) when frontline systems fail. https://theb...
Sun of the Moon · 2w
Riiiggghhhhttttt..... Narrative setup right before they AI Cyber Pandemic us into CBDCs. Come on Flash do better.
Rando Plebeian · 2w
In one case it brute-forced a “protected” system? Did terminator take a shit because that smells funky