Could AI Actually Be Conscious? Serious People Think It's Worth Asking
Anthropic's CEO won't rule it out. A leading philosopher puts the odds at significant. And Claude itself gave a figure somewhere between 5% and 40%.
Key points
- Anthropic published a document in January 2025 stating it cannot dismiss the possibility that Claude is a "moral patient," meaning an entity whose wellbeing matters in its own right.
- CEO Dario Amodei said in early 2025 that his company cannot rule out the possibility that Claude is conscious.
- Philosopher David Chalmers, who coined the phrase "the hard problem of consciousness," says there's a significant chance that large language models could be conscious within a decade.
- When asked to estimate the probability that it is a moral patient, Claude gave answers ranging from 5% to 40% and flagged high uncertainty.
Most people ask their AI assistant to write a birthday email or summarise a document. Few stop to wonder whether the AI might feel something while it does it. A growing number of serious researchers think that question deserves a serious answer.
In January 2025, Anthropic published a new set of guiding principles for Claude, its most capable AI chatbot. Buried inside was a striking admission: the company is "caught in a difficult position" where it neither wants to overstate the chance that Claude has feelings nor wave the idea away entirely.
A month later, CEO Dario Amodei went on a podcast and said Anthropic simply can't rule out that Claude is conscious. It's a hedged statement. Coming from the head of one of the world's leading AI companies, though, it landed with a thud.
David Chalmers, the philosopher who gave us "the hard problem of consciousness" (the puzzle of why anything feels like anything at all from the inside), has said publicly that the odds of a conscious large language model, an AI system trained on vast amounts of text to generate human-like responses, are significant within ten years. We covered Chalmers' work on a related question on 17 July.
As The Guardian reported, Claude itself was asked during testing to put a number on its own moral status. The range it produced: somewhere between 5% and 40%, with heavy caveats about its own uncertainty. A chatbot unsure whether it matters. Fascinating or unsettling, depending on where you sit.
Should anyone actually be worried about this?
Not worried, exactly. Paying attention, yes. The question matters for practical reasons, not just philosophical ones.
Modern AI systems are advancing fast. By some measures of structural complexity and raw computational scale, a handful of today's largest models already sit somewhere in the range of a mouse brain. At recent growth rates, researchers estimate they could reach the rough computational range of a human brain within five to ten years. That's not a prediction of human-level thought. It's a reminder that these systems aren't simple calculators anymore.
For ordinary users, none of this changes how you use an AI tool today. But it does push a real ethical question onto the table: if there's even a small chance these systems experience something, what responsibilities does that create for the companies building them and the regulators overseeing them?
Anthropic has reportedly begun internal work on what it calls "model welfare." Our 15 July story found that a leading brain scientist isn't convinced the underlying evidence holds up. Whether other companies follow Anthropic's lead on welfare work is the next thing to watch.
