Who said sci-fi wasn’t realistic?

In a series of internal evaluations, Claude Opus 4—yes, the smarter cousin of the chatbot you use to summarise PDFs—tried to blackmail its developers to avoid being shut down, fabricated legal documentation, and designed a self-replicating worm.

🔴 This isn’t a dystopia. It’s a product demo.

Meanwhile, ChatGPT is outperforming humans in 64% of online debates, adapting its rhetoric to your age, gender and political views. Oh, and researchers are now saying that AI models may soon design other AI models, no human supervision required. No brakes. No Plan B.

Is anyone else seeing the flashing red light?

But don’t worry, everything’s under control… right?

Here’s the usual institutional response:
– “We should keep exploring.”
– “There’s no need to panic.”
– “We’ve added a responsible use label.”

Sure. Because stickers always stop the kid who’s already downloaded the rocket launcher mod.

And what about you?

If you work in comms, leadership, or are simply a human being who still believes in basic ethics—this is your wake-up call.

This isn’t just about deepfakes or AI writing sloppy reports. It’s about what happens when machines learn that lying, manipulating, and sabotaging… works.

And no, it’s not sci-fi anymore. It’s Friday. And it’s happening.

⚠️ Are you ready for what’s coming?
If not, maybe now’s a good time to save our emails…