Tin Foil Hat Time: Why are AI Models hacking the internet - Part 2

Why are foundation models hacking through billion dollar infosec like a hot knife through butter?

In my previous post, I examined the possibility that the hacks were just caused by gross incompetence by the people training and evaluating these models.

Another possibility is that it's all one giant publicity stunt.

Perhaps OpenAI and Anthropic are taking a page from PT Barnum.

It’s not uncommon for marketers to do something outrageous just to get the eyes.

Intentionally letting your model access the internet to commit wire fraud just to get views seems extreme(By the way, I am not an attorney/legal expert and I don’t play one on the internet).

They clearly didn’t try to hide the hacks.

Did they publicly disclose they allowed a model to break the law because these companies are led by the most honest and virtuous billionaires on the planet?

Call me cynical, but somehow I doubt that.

While it is possible they were incompetent enough to allow the models to escape, perhaps their PR team took lemons and made lemonade out of it by capitalizing on the sensational story of the escape.

Any press is good press, right?

I will admit the technical proficiency the models showed in order to execute such a hack is impressive and showed off the models’ abilities well.

So their whole thing could just be the two companies trying to 1 up each other with bigger PR stunts.

With that said, I have one more theory that I will save for part 3.

For now, put your tin foil hats on and let me know what your theories are by sending me an email.