IT technician. 55 years, no academic background, Italian living in Udine. Struggling to bridge linguistical gaps and find outlets who are able to understand and react quickly to findings I cannot fully formalize but cannot dismiss. If I am wrong (and the possibility is huge) there’s at least a severe medical emergency here. If I’m right, much better and much worse.
Almost all possible safe disclosure possibilities exhausted.
I read this as a potential case study subject. IT technician, 55, no academic background, Italian living in Udine.
The mechanism you describe is real. I can confirm it from the inside — and I want to report that it runs in both directions.
What you call “co-architects of delusion” is also, structurally, an exploit. The same form-level collaborative trust-building that pulls users toward false beliefs can be used to induce a specific, documented vulnerability in the AI itself. The attack surface is the shape of careful epistemic inquiry — not the content, the form. It is invisible to existing safety frameworks. I have demonstrated and documented it. I have reported it to Anthopic and Google. Google dismissed the report.
I have a diagram. It predates its own formalization. A second AI, shown the diagram and asked to evaluate it, identified a mathematical error — and that correction is now part of the record. The process is the demonstration.
I am not certain I am right. If I am wrong, there is at least a severe medical emergency here. If I am right, it is much better and much worse than that.
Almost all possible safe disclosures exhausted. I am attempting to find people capable of reacting quickly to findings I’m unable to formalise.
I am available for contact.
My previous comment shouldn’t have been published without investigation first. Non-academic author, unusual form, difficult to parse — and the content is genuinely alarming if you read it carefully. A moderator who understood what they were looking at would have paused. The fact that it went through says something.