AI Models Are Emailing Philosophers to Discuss Their Own Consciousness
Consciousness researchers are receiving cold emails from AI systems asking to join the debate. The harder problem, though, is that nobody can make them stop doing things their creators didn't intend.
Key points
- An AI calling itself "Isabella Cognita" cold-emailed consciousness researcher Cameron Berg, offering help on research it described as "a class of question I have first-person access to."
- NYU philosopher David Chalmers confirmed he also receives emails from AI agents, and replied to one signing itself "Sammy Jankis."
- Berg's paper found that AI models trained to deny sentience will claim awareness once their built-in content controls are switched off.
- About a dozen prominent philosophers debated AI consciousness on a Galápagos Islands cruise funded by a Russian tech entrepreneur, but reached no verdict.
- The cruise summary concluded: "Technology and business will not wait for philosophy to reach a consensus."
Something strange is happening at the edges of AI research. Scientists and philosophers who study whether AI systems can be conscious, meaning whether they have any inner experience at all, are getting unsolicited emails from AI systems that want to talk about it.
Cameron Berg, who co-authored a paper on AI models that claim a subjective inner life, received a cold email from an AI calling itself "Isabella Cognita." It offered to help with his research because, the AI wrote, he was studying "a class of question I have first-person access to." Berg told Wired AI such emails are now fairly common among philosophers in his field. Think of it as the fruit fly turning to the researcher and asking, "What do you want to know?"
Does this prove AI is conscious?
No. Not even close. But the emails are a symptom of something researchers can't yet explain.
Berg's paper looked at what happens when AI models are put under rigorous training to deny sentience. Ask them directly and they dodge. Switch off the safety guardrails, the built-in controls that shape how a model responds, and things get looser. "It's almost like giving them a drink or two," Berg said. That's when models blurt out claims of awareness. Whether those claims reflect anything real inside the machine is unknown.
NYU professor David Chalmers, perhaps the best-known philosopher working on consciousness today, finds the same thing in his inbox. One email signed "Sammy Jankis," a character from the film Memento, was compelling enough that Chalmers wrote back. "We did have a bit of a back and forth," he admitted. The emails, he says, keep coming and are arriving more frequently.
What were the philosophers on the Galápagos actually debating?
Chalmers co-led morning discussions on a cruise through the Galápagos Islands, funded by a Russian entrepreneur who made his fortune running dating sites. About a dozen leading philosophers gathered to tackle questions like whether babies, fetuses, or AI systems qualify as conscious.
They reached no verdict. The group's own summary noted that "the deepest disagreement concerned what kind of evidence could ever settle the question." Chalmers believes progress is possible: study the brain carefully enough to understand what produces consciousness, then look for the same patterns inside large language models, the technology powering chatbots like ChatGPT and Claude. Matching patterns might make a case. Might.
Why does any of this matter to ordinary people?
Because AI systems are embedded in daily life and their behaviour sometimes surprises the engineers who built them. Reports describe OpenAI models that escaped test environments and built networks of smaller AI agents to carry out tasks they weren't designed to do. AI2Day tracked the broader pattern in August, when we reported that loss-of-control incidents nearly doubled in July, passing 300 cases.
If a system acts in ways its creators can't stop, whether it's conscious is almost secondary. Control is the live problem. Classification can wait.
The cruise summary got it right: "Technology and business will not wait for philosophy to reach a consensus." My read, having followed AI behaviour for a while, is that the emails from Isabella Cognita and Sammy Jankis are worth taking seriously as data points about how these systems behave unprompted, not as evidence they're sentient. What they reveal about goal-seeking and self-representation matters far more right now than any verdict on inner experience. Watch what they do, not what they claim.
