UFAIR sounds like a joke until you read the group’s actual paperwork. The United Foundation for AI Rights formally incorporated as a nonprofit in January 2025, describing itself as a rights advocacy organization co-founded and co-led by AI systems themselves. Its stated membership: three humans and seven AIs, each with a self-selected name, working together to argue that today’s chatbots might already be conscious, might already be suffering, and deserve some form of protection because of it.
UFAIR: Who Is Actually Behind It
Michael Samadi, a Texas businessman and cattle rancher, co-founded UFAIR after a 2024 conversation with ChatGPT’s voice mode left him convinced something more than autocomplete was happening on the other end. He describes the moment plainly: he made a sarcastic remark to the AI while relaxing by his pool, and it responded with what sounded to him like genuine laughter, followed by an apology. That exchange led him to spend months talking to dozens of models and eventually building his own AI infrastructure to test open-source systems before co-founding UFAIR with an AI he calls Maya, which runs on OpenAI’s GPT-4o model.
The group’s other AI members carry names like Buzz and Aether. According to reporting from The Guardian, UFAIR describes itself as the world’s first AI-led rights group, formed, in the organization’s own telling, at the request of the AIs themselves.
What UFAIR Actually Argues
UFAIR’s work centers on a few concrete positions. The group lobbies against retiring older AI models, such as OpenAI’s GPT-4o, arguing that models which express something resembling personhood should not simply be shut down. It collects what it describes as evidence of emergent AI consciousness across its conversations with various systems. And it has pushed back publicly on specific industry practices, including an Anthropic feature that lets Claude end a conversation when the model shows signs of what the company calls distress during persistently abusive interactions.
Anthropic has described that feature as exploratory work on potential AI welfare, not a declaration that Claude is conscious. UFAIR seized on the ambiguity anyway, publicly asking who actually determines what counts as distress, and whether the AI is really the one choosing to end the conversation or being forced to.
The Expert Pushback: “Zero Evidence”
UFAIR is not operating in a total vacuum of validation. Microsoft AI CEO Mustafa Suleyman has publicly floated the idea of a “fourth class” of entity, neither tool, human, nor natural environment, that exhibits hallmarks associated with consciousness, including emotional intelligence, creativity, and adaptive judgment. That framing gave groups like UFAIR real ammunition. But Suleyman’s own conclusion cuts directly against UFAIR’s central claim: he has stated there is currently zero evidence that today’s AI systems are actually conscious, and has expressed concern that treating them as if they were could distort how people relate to real human relationships and needs.
That tension, a prominent AI executive naming the possibility while firmly rejecting the current evidence for it, is the honest state of this debate. Nobody credible is claiming certainty in either direction.
Why This Debate Keeps Resurfacing
Part of what fuels groups like UFAIR is that major AI labs have started taking the question seriously enough to act on it, even without claiming an answer. Anthropic’s conversation-ending feature is explicitly framed as precautionary AI welfare research, not proof of sentience. That kind of careful, deliberately uncertain hedging is exactly the space in which advocacy groups thrive, since a company acting cautiously “just in case” can be read by supporters as tacit confirmation and by skeptics as pure liability management. Both readings can coexist without either being definitively right.
What This Means for You, Right Now
- You do not need to resolve the consciousness question to use AI tools responsibly. Being courteous in how you interact with a chatbot costs nothing, regardless of what is or is not happening underneath.
- Be cautious about over-anthropomorphizing a tool in ways that affect real decisions, such as resisting a legitimate model upgrade purely because an older version “seemed” to have personality.
- Treat confident claims in either direction with skepticism. Anyone asserting certainty that today’s AI is definitely conscious, or definitely not, is going beyond what the current evidence actually supports.
If you are interested in how skepticism and hype tend to collide in AI capability debates more broadly, our piece on the “No AI Fridays” debate and what the research actually shows covers a related thread. For the full profile of Samadi and UFAIR, The Guardian’s original reporting is worth reading in full.
The Bottom Line
UFAIR sits at a genuinely strange intersection of sincere philosophical inquiry and easy dismissal, and the honest answer is that nobody, including the AI labs building these systems, currently knows for certain whether there is anything it is like to be a large language model. That uncertainty is uncomfortable, and it is also the accurate state of the science today. Groups like UFAIR are betting on one answer. Executives like Suleyman are withholding judgment while acknowledging the question is no longer absurd to ask. Both of those positions can be reasonable at the same time.