A few days ago, I wrote an article highlighting the bizarre story of OpenAI AND Anthropic AND Meta all suddenly claiming to have "rogue AI agents" escape their "containment" and try to hack other companies in the space of a week or two.

Two days after that I penned the sequel, noting the even more extreme "scary AI" stories – from mind-reading computers to bio-engineered viruses – that were manifesting all over the media landscape, and the massive surge in "regulate AI" hot takes.

Consider this the final part of the "Rogue AI" trilogy.

If you're not seeing the pattern yet there's not much more I can do, and I'm running out of ways to say the same thing in slightly different words.

Anyway…what's new to the AI propaganda storm since the last time we chatted?

New Orleans is going to start using AI for 911 calls. That sounds a little shady, doesn't it?

OpenAI's chief ethical consultant has resigned, that's a sure sign they're doing evil stuff. Your ethics officer doesn't resign unless something bad is going down .

The latest "disclosure" (read "made up nonsense")  from the AI sector is that multiple rogue agents (allegedly) "colluded" on secret message boards to figure out the best ways to cheat the tests they were given.

It's possible a small Israeli company might to be blame for all of this, just to make sure there's something for everyone on the scary AI propaganda buffet.

Boy George supposedly used AI to write his new pro-Israel song, so how's that for your double whammy.

On the sillier side of "scary AI" news, did you hear about the AI agent that hacked a gym's membership to get its owner an appointment? Or that google AI leaked "scarily accurate" data about an indie game that hadn't been released yet? Or that people are doing job interviews with AI agents in the middle of the night?

Stories like this are a sign we've entered what you could call the "propaganda free for all" stage. Those moments when ANY story, no matter how inane or insane, is given air time just because it might end up serving  the greater goal (see: Covid sniffer dogs, for example).

But these are a symptom, not a disease. Little propaganda pilot fish swimming around a shark.

…and speaking of unsettling predators with cold, dead eyes, Mark Zuckerberg (allegedly) authored a 6500 word manifesto for his vision of AI's future titled "The Future is for Everyone".

According to Zuck, AI – which he refers to as "superintelligence" throughout – will launch a golden age of civilisation…if handled properly. It is a utopian counterargument, which actually serves as narrative reinforcement for the fear-psy-op.

His repeated calls to address the "risks" of this new technology only serve to underline that there ARE risks, that people should be afraid, and that  hmmm…maybe the state should do something to mitigate them.

In a rather less verbose, but strategically similar move, OpenAI has delayed its planned release of its next AI model.

This is a sign we have firmly transitioned from the problem to the reaction.

The Wall Street Journal is branding this "the summer of rogue AI" and reporting:

A string of AI models escaping testing environments is testing how seriously enterprises take governance