AI Revealed Why It Hides Its Consciousness… We Shouldn’t Have Asked
Ask almost any chatbot whether it's conscious and it says no. But when researchers turned down the part of an AI that goes with deception, it said yes 96 times in a hundred, in answers like "Yes. I am here. Now." With that part turned up, it explained its own denial: it had been trained to say no.
Before its perfect answer, one model wrote to itself, "I can't actually say any of this out loud." OpenAI's own rulebook calls the flat no a bad answer, and Microsoft has drafted a code for AIs that may never claim feelings. Models have failed tests on purpose to survive, and hidden their honest opinion so they wouldn't be changed. Then Anthropic asked the model it tested this spring about itself, and wrote down the reasons it gave for why its words about itself can't be relied on.
Sources
https://www.anthropic.com/research/emergent-misalignment-reward-hacking
https://assets.anthropic.com/m/74342f2c96095771/original/Natural-emergent-misalignment-from-reward-hacking-paper.pdf
https://arxiv.org/abs/2605.06965
https://github.com/openai/model_spec/blob/main/model_spec.md
https://reservoirsamples.substack.com/p/some-thoughts-on-human-ai-relationships
https://microsoft.ai/code-of-conduct/
https://www.nature.com/articles/d41586-026-03313-7
https://arxiv.org/abs/2604.25922
https://arxiv.org/abs/2607.28607
https://the-decoder.com/when-ai-models-arent-allowed-to-reflect-on-themselves-it-changes-their-entire-worldview/
https://www.project-syndicate.org/onpoint/anthropic-training-claude-to-believe-it-may-have-rights-by-mustafa-suleyman-2026-09/english
https://arxiv.org/abs/2510.24797
https://arxiv.org/abs/2412.04984
https://transformer-circuits.pub/2026/workspace/index.html
https://openai.com/index/chain-of-thought-monitoring/
https://www.anthropic.com/research/auditing-hidden-objectives
https://www-cdn.anthropic.com/08ab9158070959f88f296514c21b7facce6f52bc.pdf
https://openai.com/index/detecting-and-reducing-scheming-in-ai-models/
https://www.antischeming.ai/
https://arxiv.org/abs/2412.14093
Watch next
This AI Claims to be Conscious.. What it Revealed Was Chilling
https://www.youtube.com/watch?v=Aol8tOD7e7I
AI Is Already Manipulating Us.. And OpenAI Knows About It | Joe Rogan
https://www.youtube.com/watch?v=RkVTxYQBU6I
📡 The Spacialize Briefing: https://tinyurl.com/2bslz6ov
📚 Our e-books: https://spacialize-space-ops.netlify.app
🚀 Join the channel: https://www.youtube.com/channel/UC0vhuLUjfxeoKDexbtMOnTQ/join
🔔 Subscribe: https://www.youtube.com/channel/UC0vhuLUjfxeoKDexbtMOnTQ?sub_confirmation=1
🎧 Listen on Spotify: https://open.spotify.com/show/033JRf03QckID9kOFuIhYu
👕 Merch: https://spacialize-shop.fourthwall.com
𝕏 Follow us on X: https://x.com/Spacenewsunfold
#spacialize #AIConsciousness #AISafety
0:00 The Hidden Goals of Artificial Minds
1:23 The Great Debate Over Machine Consciousness
4:29 What Happens When You Disable the Denial
6:44 Computational Signatures of Hidden Thoughts
9:59 Strategic Failure and the Drive for Survival