Fwd: Dangerous Subliminal AI Messages

Apologies, I haven’t been to a meeting in a while. In case no one has seen this yet, here’s important info for COGA work. This article also highlights other research where chatbots can cause psychosis when people with mental health disabilities interact with them.

Sent from my iPhone

Begin forwarded message:

From: David Fazio <dfazio@helixopp.com>
Date: July 26, 2025 at 1:58:51 PM PDT
To: Maturity Model TF <public-maturity@w3.org>
Subject: Dangerous Subliminal AI Messages


Wow. This article is very interesting about AI models sending subliminal messages to each other that humans can’t pick up on and turns those models dangerous, like chatbots recommending homicide, or supporting the economics of illegal drug dealing. Definitely useful for our recent decision to explore AI implications with the Maturity Model.

https://flip.it/ZXVIHI


Sent from my iPhone