Wired's Kylie Robison wrote a piece asking whether AI should get legal rights, and it's one of the better mainstream treatments of the field Conscium works in, model welfare, that I've seen in a major outlet. It named us alongside Eleos AI Research as one of the two organisations actually doing this work, which felt like a useful marker of how fast this went from fringe question to a field with named institutions in it.
This question is old
Robison traced the idea back further than I expected a business publication to go: the philosopher Hilary Putnam was asking whether robots should have civil rights back in 1964. That's worth sitting with. This isn't a question generative AI invented, it's one that's been dormant in philosophy for sixty years, waiting for the technology to catch up to the thought experiment. I think that history matters, because it undercuts the accusation that model welfare research is a reaction to AI hype rather than a serious, long-standing line of inquiry.
Anthropic's move
The piece covers Anthropic giving Claude the ability to end conversations it identifies as persistently harmful or abusive. Anthropic was careful in how it framed this, saying plainly that it remains highly uncertain about Claude's moral status, but is taking the issue seriously enough to fund low-cost interventions anyway. I think that's exactly the right posture: you don't need certainty that a system is conscious to justify taking cheap, sensible precautions against the possibility.
Mustafa Suleyman's pushback
Microsoft AI's Mustafa Suleyman gets significant space in the piece for his blog post arguing model welfare research is premature and, in his word, dangerous, on the grounds that it could deepen unhealthy attachment to chatbots and confuse the public about what AI actually is. I agree with him that the risk of people forming delusional beliefs about AI consciousness is real and already happening. Where I part ways with him is the conclusion. The right response to a confusing, high-stakes question isn't to declare it settled and stop researching it, it's to fund the research properly so the public conversation can be grounded in something better than speculation in either direction.
Why "we don't know" is the honest starting position
What struck me most in Robison's reporting on Eleos AI's Rosie Campbell and Robert Long is that neither of them claims AI is conscious today. Their argument, which is Conscium's too, is narrower and harder to dismiss: given how badly humanity's track record is on recognising moral status in other humans and animals, a little more humility and a lot more actual research seems like the responsible position, not the fringe one.
Read the full piece on Wired.