Schneier on Security

Clive Robinson • September 14, 2025 12:40 AM

@ Bruce,

I forgot to ask earlier,

“Are you an Eagle or Owl fan?”

An odd sounding question perhaps, but it is in effect from an interesting video on the Subliminal message issue of,

“These Numbers Can Make AI Dangerous”

https://m.youtube.com/watch?v=NUAb6zHXqdI

The papers involved are from after you started your book. But the precise mathematical details are just an instance of a more general issue.

The net result is that it appears that there are hidden channels that can be used by an attacker to bias thus pass on a prejudice, or political mantra to more than just an arms length.

The implication, for both the generation and use of legislation by AI, is actually something people need to think about.