32 - Regulatory Propaganda

This one's maybe a little long, but an attacker stole a Meta API key, which used $600,000 worth of API credits, and no one noticed for weeks.

What? Yeah.

So AI model testing organization METR has disclosed two attacks that happened earlier this year, including one in which an attacker stole an API key, then spent three weeks consuming public model credits worth about $600,000.

Now, METR in this case is short for Model Evaluation and Threat Research, found no evidence that the attackers accessed sensitive information in either incident, and the org said it is investigated both with security experts.

Now, and I don't know that these are actual security experts or not.

I was going to say

we know what experts are.

We know what

experts are, but there are real experts.

We've actually encountered a couple this week in a completely different way.

Yeah, but they

were actually competent.

They were

actually competent, yes.

I'm right here.

Well, you too.

So meta researchers worked with OpenAI to investigate how its agents attacked Hugging Face, and on Monday it disclosed two of its own security snafus.

Yes, in the security incidents, you know.

How many more times does OpenAI have to cock things up before there's any degree of accountability with these agents? Oh, no, this is the security

experts.

This is not...

I mean, this is about the hugging face thing, right? So this is a previous snafu by OpenAI, but this is the company that they use to help sort of, I guess, help to figure this out or something.

You know, they're supposedly experts, yet they have their own security issues.

Yeah, METR is a security firm.

At the risk of unburying the lead here, OpenAI...