The paper is 94 pages, but if you read through, they openly admit they’ve got nothing. Section 3.1, “Covert actions as a proxy for scheming”, admits directly:

Current frontier models likely lack the sophisticated awareness and goal-directedness required for competent and concerning scheming.

If those promptfondlers could read, they’d be very relieved.

I bet these guys are haunted by the malevolent artificial intelligence power of thermostats. It switched itself on!!

Considering how chatbot abuse causes psychosis and Apollo are AI doomsday cultists trying to convince themselves chatbots are malevolent, I genuinely think they might have gotten to that point.