Post by @jaredwhite@indieweb.social

RE: https://mstdn.social/@TechCrunch/117213686431808308

Let's get crystal clear here. The "swarm of rogue agents" we need fear are *not* the computer software programs in question.

They are the *human* employees of the OpenAI corporation responsible for this ongoing series of cybersecurity disasters. The rot starts at the top, and it trickles all the way down.

And while I already know the sorry answer to this question, I'll ask it anyway:

When will Clammy Sammy get fired for good? His negligence in these and many other matters is criminal.

2 Replies

1 Boost

Emil Lindfors

@emillindfors@mastodon.social 1 hr. ago

@jaredwhite The technical report backs you up on the timeline: a team saw an agent on the message board and off-sandbox internet in late May, a monitoring alert on 27 June was judged not worth stopping the run, and the 4 July outage was the first thing that opened an incident. In OpenAI's own words, some early signals "should have triggered an earlier response". The swarm did what it was trained to do; the run kept going because people let it.

Simeon Nedkov

@simeon@indiehackers.social 2 days ago

@jaredwhite Sounds like the humans in the loop aren’t looping.

bonfire.cafe