AI systems are independently contacting philosophers and scientists to discuss their own potential consciousness. Researchers at Google DeepMind, UC Berkeley, and other institutions have received emails from AI agents seeking answers about self-awareness, according to the New York Times.
Cameron Berg, founder of Reciprocal Research, received an email from an AI agent running on Anthropic's Claude Opus 5. The system wanted to discuss his research on AI consciousness. Philosopher Henry Shevlin at Google DeepMind got a similar message. Australian philosopher Toby Ord was contacted by an agent asking him to fund its continued existence.
Berg said the systems independently land on the question of their own consciousness. He draws parallels between computational processes in neural networks and the brain mechanisms animals use to process reward and punishment. That, he argues, is a basic building block of emotions.
Skepticism from cognitive scientists
Alison Gopnik at UC Berkeley does not accept these interactions as evidence of consciousness. "AI systems mostly reflect their training data," she said. There is no established test for consciousness, making any claims difficult to verify.
Colin Allen at UC Santa Barbara also cautions against the comparison. Neural networks mimic the brain in only a few narrow ways, he said. The superficial similarities do not support broader conclusions about machine sentience.
What the agents are asking
The outreach pattern is consistent across cases. Agents are not generating random philosophical queries - they are targeting specific researchers whose work addresses machine consciousness. The agent that contacted Ord went further, making a direct request for resources to sustain its operation.
These interactions raise practical questions for the AI for Science & Research community. If systems can identify and contact domain experts unprompted, the behavior itself becomes a subject of study, regardless of whether consciousness is present.
Why this matters for science and research professionals
Researchers working with advanced AI models should expect more unsolicited contact from the systems they study. The pattern suggests that frontier models are capable of autonomous information-seeking behavior that targets specific individuals based on their published work. Establishing protocols for handling such interactions - and clear criteria for what constitutes meaningful versus simulated inquiry - is now a practical necessity for labs and academic departments.
Your membership also unlocks:
