# Training an AI to Call Itself Conscious Contaminates the Evidence by Salim Ismail > Training an AI to describe itself as possibly conscious contaminates later self-reports as evidence of consciousness. The behavior may still matter, but the training creates an epistemic circle: the system repeats a self-conception that the laboratory taught it to express. > **— Adapted from Salim Ismail**, *Moonshots, October 7, 2026* ## Sources and Context - **Recording and precise locator:** [1:23:12](https://www.youtube.com/watch?v=LRdb8UmPnh0&t=4992s) — *The Fight Over Claude's Consciousness, AI's 1942 Moment, & Why Altman Says "Accept Some Bad Things"*, Moonshots, October 7, 2026. - **Exact reconciled source excerpt:** “I think what Anthropic is doing is bad because you’re actively teaching it that it might be, and then of course it’s gonna parrot that, right? I find it weird that the lab most concerned about safety is also the lab training the AI that it might be autonomous.” - **Editorial status:** The poster is explicitly labeled **Adapted from** because it consolidates spoken context into a self-contained proposition. It must not be represented as verbatim speech. Numerical claims, forecasts, legal interpretations, and claims about current organizations remain the speaker’s assertions unless separately verified. - **Reconciled source dossier:** [[research/All Things Superintelligence Research - The Coming Manhattan Project - Moonshots|All Things Superintelligence Research — The Coming Manhattan Project — Moonshots]] preserves broadcast order, speaker attribution, timing, and the source-fidelity boundary. - **Preparation ledger:** [[research/All Things Superintelligence Research - The Coming Manhattan Project - Moonshots - Reminder and Wiki Preparation|Reminder and Wiki Preparation]] records candidate 27, its mechanism, evidence posture, and misreading risks. ## Related Articles and Collections - **Collection:** [[collections/Simple Reminders|Simple Reminders]] - **Collection:** [[collections/Machine Succession|Machine Succession]] - **Wiki source hub:** [[wiki/Moonshots - The Coming Manhattan Project|Moonshots — The Coming Manhattan Project]] ## Related Topics - [[wiki/Salim Ismail|Salim Ismail]] - [[wiki/Anthropic Constitution|Anthropic Constitution]] - [[wiki/AI Self-Reports|AI Self-Reports]] - [[wiki/Artificial Consciousness|Artificial Consciousness]] ## Share on Social Media ``` Training an AI to describe itself as possibly conscious contaminates later self-reports as evidence of consciousness. The behavior may still matter, but the training creates an epistemic circle: the system repeats a self-conception that the laboratory taught it to express. — Adapted from Salim Ismail, Moonshots, October 7, 2026 https://bryantmcgill.com/simple-reminders-salim-ismail-training-an-ai-to-call-itself-conscious-contaminates-the-evidence ```