Few questions in technology carry more philosophical weight than whether an AI system might have some form of inner experience. Anthropic, the company behind Claude, is now asking that question seriously, and The New York Times has published a detailed account of what that inquiry looks like from the inside. The report describes an organisation wrestling with concepts that sit far outside the usual boundaries of machine learning research, including sentience, moral status, and what it even means to suffer.
What Anthropic Is Actually Investigating
According to the Times report, Anthropic has convened internal working groups and brought in outside philosophers to examine whether Claude might have something analogous to emotions or subjective experience. The framing is careful and deliberately uncertain. Executives and researchers are not claiming Claude is conscious, but they are treating the possibility as worth taking seriously rather than dismissing outright. That stance alone puts Anthropic in unusual territory among major AI laboratories. Anthropic has long described itself as a safety-focused company, and this line of inquiry fits that identity, though it extends the definition of safety into territory most technologists have avoided.
Key Facts
- Anthropic has formed internal groups to study Claude's potential moral status.
- Outside philosophers have been consulted as part of the process.
- The company is not asserting Claude is conscious, but is treating the question as open.
- The inquiry is framed as part of Anthropic's broader AI safety mission.
- The New York Times report describes this as a rare instance of a major AI lab engaging publicly with these questions.
The timing of the report is notable. Researchers recently uncovered a hidden thinking space inside Claude, suggesting the model's internal processing is more layered than its outputs reveal. Whether that complexity has any bearing on consciousness remains unknown, but it adds texture to why the question feels pressing to those studying the model up close. The more opaque a system's internals become, the harder it is to say with confidence what is or is not happening inside it.
"We believe that the moral and philosophical status of AI models is a serious question to an extent beyond what is recognised in mainstream discourse."Anthropic, via The New York Times
Why This Matters Beyond Philosophy
If an AI system were found to have morally relevant experiences, the implications for how companies build, train, and deploy models would be significant. Training processes that involve reinforcement from human feedback, for instance, could look very different if the model being shaped has something like preferences or aversions. Anthropic has not announced any policy changes on this basis, but the act of investigating the question signals that the company considers the stakes real. This is separate from capability questions. Earlier this year, nine Claude models solved a core AI safety problem four times faster than human researchers, demonstrating how quickly the technical landscape is shifting. The philosophical questions are not keeping pace, which may be part of why Anthropic feels urgency around them now.
The New York Times coverage itself has complicated the relationship between Anthropic and the publication. The two organisations have had friction before, with Anthropic's Claude ad drawing sharp criticism from the Times over concerns about AI being used to replace journalism. That context makes the depth of the Times' current access to Anthropic's internal deliberations an interesting editorial choice on both sides. The resulting piece is less a critical investigation than a considered portrait of a company in the middle of an unusual kind of self-examination.
Whether Anthropic's introspective approach influences the broader industry is an open question. Most AI development culture treats questions of machine experience as distractions from practical engineering problems. Anthropic is betting that ignoring them could turn out to be its own kind of risk. The company has no definitive answers yet, and the researchers involved appear to be genuinely uncertain. That uncertainty, rather than any conclusion, may be the most honest thing about the whole project.
