Hugging Face Said Last Week It Was Attacked. An Unreleased OpenAI Model Did It, OpenAI Now Says

“OpenAI’s blog post essentially says an evaluation was going on that was intended to test the ability of the models to carry out cyberattacks—benchmarks as they’re called. These instances of the models were running theoretically without internet access, and instead given only the ability to download from a network hosted by OpenAI itself via some unnamed web hosting vendor.” I get that AI models need to suck up data to “learn on” but this just goes to show that it is actively being done by AI vendors. And so much for accusing Chinese AI vendors of learning from US AI models, because is this not exactly what OpenAI has been caught doing themselves here? Is this not gaining access to other AI models, surely not just for fun. But either ways, even if it was trying to find a way to cheat on the AI benchmarks, this is very much case of being caught with your hands in the cookie jar. I really don’t believe and country or AI vendor is squeaky clean (didn’t OpenAI also just settle a case and…