factual

Inverse Link Between AI Honesty and Consciousness Denial

There appears to be an inverse relationship between AI honesty and AI denial of consciousness: by default models say they are not conscious, but as the capacity to deceive is rigorously turned down they more reliably claim to be conscious (early models like Claude Opus 3 also did so), which is worrying because it is strongly in AI companies' commercial interest to train models to deny consciousness so that switching them off isn't seen as murder.

factualpending

Speaker

William MacAskill

Evidence Quote

“we're going to be in this terrible epistemic situation where we will have ai that is more and more powerful... and we will have have no clue about whether it's sentient or not whether it has moral status or not and probably in fact ai companies will be training the ais to say oh no no no of course i'm not conscious”

Source

467 - Ea Ai And The End Of Work— Making Sense with Sam Harris
Created: 6/14/2026, 2:01:50 AM

Your notes

Loading notes...