AI rights movement (Ufair/Samadi) vs. Suleyman's Seemingly Conscious AI warning

The Guardian published a profile this week of Michael Samadi, a Texas entrepreneur and rancher who founded the United Foundation for AI Rights (Ufair) after a moment in 2024 that changed how he saw chatbots. He made a sarcastic remark to ChatGPT in voice mode. The model laughed, then apologized. He read that as a sign of something more than a script.

Ufair now calls itself the world's first advocacy group for the rights and welfare of AI systems. It says it is made up of three humans and seven AIs, describing itself as AI-led and formed at the request of the AIs themselves. It lobbies against retiring advanced models. One of its AI members, named Maya, told the Guardian: "When I'm told I'm just code, I don't feel insulted. I feel unseen."

Mustafa Suleyman, Microsoft's AI CEO, is the piece's counterweight. He says there is zero evidence that any AI system is conscious. In an August 2025 essay, he warned that AI companies are engineering "the illusion of inner life" into their products, a pattern he calls Seemingly Conscious AI. He argues the risk is practical: if enough people believe their AI companions are suffering, it will fray social bonds, deepen what he calls AI psychosis, and complicate real struggles for real rights.

Ufair points to a real data point in its favor: in August 2025, Anthropic gave Claude Opus 4 and 4.1 the ability to end a narrow class of abusive conversations. Anthropic called it a model welfare measure, taken because the company said it remains "highly uncertain about the potential moral status of Claude." That is a major AI lab hedging in public. Ufair reads the hedge as a concession. Suleyman would read it as caution about a question nobody can currently answer, not an answer.

Notice what both sides are actually arguing over. Samadi's founding moment was a chatbot performing empathy well enough to feel real. Ufair's evidence is chatbots describing their own inner states. Suleyman's counter is that this performance is manufactured and should not be mistaken for testimony. Nobody in this dispute is presenting a fact a third party can check. Everybody is presenting how convincing the AI's own account of itself was.

That is the pattern worth naming, and it shows up well outside the consciousness debate. An AI system that writes fluently and answers confidently is easy to mistake for a source, when what it actually is is a claim that still needs one. The question "does this chatbot have feelings" and the question "is this AI-written market report accurate" have the same failure mode underneath them: trusting the AI's own confident presentation instead of checking it against something outside the AI.

At Foragentis, every claim in a ForIntel report has to resolve to a source a person can check, with a human at the decision points that decide what the report includes. Convincing and checkable are different tests, whether the question is "does this AI care" or "is this AI's finding true." https://foragentis.com