C. Ocelot on Mythryl: “I normally don't give LLMs props but under the same testing…”

I normally don't give LLMs props but under the same testing conditions (capture the flag) by the same independent testing firm (Irregular) Gemini calculated that the data being gathered wasn't part of the game and ended operations. Same situation as the other two, Irregular didn't airgap it.

Some kind of bird-horse?!@hazelgoesneigh.bsky.social

the amount of reporting of "AI's going rogue and hacking companies in security tests" as if it's an example of some kind of emergent behaviour, rather than the AI companies playing fast and loose with test parameters is fucking exhausting holy shit. These LLMs aren't magic, they're fancy chat bots.

21/9/26, 8:24 pm0 reposts1 like

Post