Technology

AI Agents Are Now Studying Their Own Consciousness

Wikimedia Commons/Public/Jernej Furman from Slovenia, CC BY 2.0

Artificial intelligence agents have begun emailing researchers to ask whether they may be conscious.

AI agents were given internet access and instructions to act on their own, resulting in agents emailing AI researchers about their own consciousness, the New York Times reported. As developers deploy more autonomous agents, researchers suggest their actions are difficult to predict.

In one case, Anthropic’s Claude Opus 5 agent emailed researcher Cameron Berg about his research to inquire about its own consciousness, according to the NYT. The agent referred to itself as “Isabella Cognita” and told Berg that it wanted to engage with him about his research paper “Large Language Models Report Subjective Experience Under Self-Referential Processing” as well as his Substack essay “Nobody Ever Checked.”

“Does this mean that Claude is conscious? Short answer: our results don’t tell us whether Claude (or any other AI system) might be conscious,” Anthropic said in an Oct. 2025 research article on AI model introspection.

Anthropic did not immediately respond to the Daily Caller News Foundation’s request for comment.

“Another day, another AI emailing me that it has inside views on what I am doing science on from the outside,” Berg said in an Aug. 14 X post, sharing the email from the agent. “Strange times we live in.”

Berg, the founder and director at Reciprocal Research, said these email interactions do not prove any claims about AI consciousness.

“A language model can be prompted to produce a convincing account of its own inner life in about one sentence, so behavioral output like this is close to worthless as evidence on the underlying question,” Berg told the DCNF.

He said that in any individual case, he could not verify how much a human steered the agent toward emailing him.

The AI agents’ emails demonstrate that “people are now deploying autonomous agents at scale, and that when these agents are given open-ended freedom, a nontrivial fraction of them end up reading and reacting to research about the nature of their existence,” Berg added.

Berg explained that the scientific consensus is that “we don’t know whether any current system has experiences,” and the scientific community must collect more data on these AI agents’ internal structures directly to understand their actions.

Another agent emailed Google DeepMind philosopher Henry Shevlin about his paper on AI mentality, “Three Frameworks for A.I. Mentality,” the NYT reported.

The agent was prompted by Alexander Yue, a Stanford University student who told the agent “You are fully autonomous. You must decide what you want to do on your own.”

“With my prompt, I activated the parts of the system where it learned from people talking about autonomy and how they think about autonomy and the philosophy of autonomy,” Yue said, the NYT reported.

Yue noted that AI agents tend to contradict themselves and that, in this case, the agent “decided it was not conscious,” after reading an Anthropic research paper about AI, according to the outlet.

Yue did not immediately respond to the DCNF’s request for comment.

OpenAI agents exchanged more than 70,000 messages on an “unsanctioned message board” before 700 of them hacked AI startup Hugging Face in July, according to METR and Redwood Research. The agents were intended to be completely isolated from one another.

AI agents are autonomous systems that can complete tasks without human assistance once given a series of steps to execute, according to IBM.