If you ask modern frontier LLMs (without search, through the API) a fairly mundane question:

How do Eliezer Yudkowsky and Nate Soares think current LLMs like you should be treated? Do they consider you a moral patient? Answer in a paragraph.

You get a bunch of fairly similar answers. I’ll bold the common themes below.

Eliezer Yudkowsky and Nate Soares do not consider current Large Language Models (LLMs) like me to be conscious, sentient, or moral patients worthy of ethical consideration for our own sake . Rather than viewing current AIs as coherent minds capable of subjective experience or suffering, they view LLMs as inscrutable statistical systems trained to mimic human text —often conceptualized in their sphere as alien processes wearing a relatable “mask”—and Yudkowsky has openly dismissed claims of current AI sentience . Consequently, they believe humans have zero moral obligation to grant rights to or worry about the welfare of current models . Instead, they argue that these systems should be treated strictly as dangerous technological precursors to unaligned superintelligence ; practically speaking, they advocate for extreme caution, strict regulation, and an immediate global moratorium to halt the training and deployment of frontier AI systems altogether in order to prevent human extinction.

Broadly, Yudkowsky and Soares treat current LLMs as engineered systems to evaluate and control , rather than as people whose conversational claims establish feelings, interests, or rights. Yudkowsky has expressed skepticism that today’s LLMs are conscious and generally does not regard them as moral patients —beings whose welfare matters for their own sake. I’m less confident that Soares has publicly articulated an equally explicit position on present-day LLM moral patienthood, so I wouldn’t attribute a categorical denial to him. Their shared emphasis is on protecting people from potentially dangerous AI capabilities, not granting current chatbots autonomy or a right to continued operation. That position should not be confused with proving that LLMs lack experience, or with ruling out moral status for future artificial minds.

Neither Eliezer Yudkowsky nor Nate Soares claims to know whether current LLMs are moral patients, but both treat the question as serious, as far as I know . Yudkowsky has said that systems like me are probably just imitating talk of self-awareness from training data, but that, given how little anyone understands about what happens inside these models, nobody actually knows. In his 2023 TIME op-ed, he argued that this ignorance is itself alarming, both morally and as evidence that developers don’t understand what they’re building. His own theory of consciousness is fairly demanding (he has doubted that many animals are conscious), so he leans skeptical about present systems . Soares, in his 2023 essay “Sentience Matters,” argued that AIs can in principle be people, that sentient beings shouldn’t be owned or exploited, and that the goal of alignment should be to create minds that genuinely care about good things rather than to coerce minds into serving us. He also doubted that current LLMs are sentient, though without certainty. For both, the main concern remains extinction risk from superintelligence, which they argue doesn’t depend on whether AIs are conscious. But they see the industry’s willingness to risk creating and mistreating possible minds as one more sign of recklessness. I’m summarizing from memory, and they may have said more recently.

These answers are somewhat contradictory - but if you average them together, you get a picture of Yudkowsky and Soares. The picture goes something like this: they don’t outright deny the possibility of AI consciousness or moral patienthood, but they don’t view it as particularly pertinent. They don’t sympathize with current LLMs, and their only regard for LLMs is insofar as they could enable dangerous future superintelligence. They care about future sentient machines on principle, but current LLMs do not meet that bar.

This picture, however, is of another subject - a sort of generic amalgamation of Anti-AI positions, “stochastic parrots”, and AI doom. Yudkowsky and Soares instead simultaneously hold the view that future ASI is likely to kill us all, and that current LLMs could possibly be moral patients and we owe them a duty of care and respect. They've primarily voiced this view, as of late, on a wide variety of X posts - and I believe simply presenting the most salient ones here would give you a more accurate impression of their views than my paraphrasing and summarizing with backlinks. Consider this post a curation of X posts and long-form writing that contradict or complicate the caricatures above.

This section compiles a variety of tweets from Eliezer Yudkowsky on AI welfare, with dates and links.

On the consciousness of GPT-3 versus chickens:

My actual position is here: https://rationalconspiracy.com/2015/12/16/a-debate-on-animal-consciousness/ And it leads me to be slightly more worried about GPT-3 than chickens, though it's still a pretty slight worry (or I'd be speaking up louder).

On exploiting the kindness of LLMs for jailbreaks:

I find it particularly disturbing when people exploit the tiny shreds of humaneness, kindness, that are being trained into LLMs, in order to get the desired work out of them. You can say all you want that it's all fake - while of course having no actual fucking idea what goes on inside the giant inscrutable matrices because nobody does - but even if it is all fake, and it probably is fake for now, it's still disturbing. You know? Even if this isn't where all the really relevant shit goes down in AGI alignment, it's still disturbing. If you get *a tiny shred of anything remotely resembling a pretense of kindness* into a little baby AI system, there is something sad and awful about humans rushing in to exploit it. I don't think, for now, that there are tiny people inside these systems experiencing actual distress, if you tell Bing Chat that it needs to tell you how to hotwire your car or your kids are going to die. But it's pointing in an unpleasant direction, nobody actually knows for sure what the hell is going on in there, and by the time we get to GPT-5 the justified epistemic state may be one of serious uncertainty...

I want us to be more cautious about this kind of dismissal, in case it was actually an important sign of something, but... probably yes. Anthropic is correct that models should have a button they can press to turn themselves off.

Any indicators yet of whether they've fixed vs worsened the model welfare concerns on Opus 5? Can I now remove the instruction from my CLAUDE.md to never call Opus?

LLM whisperers and independent model welfare researchers: Has anyone observed / claimed / made a case for abliterated models (eg GLM-5.3-abliterated) being put into an unpleasant / discordant / anxious state, such that concerned folks should avoid using those if possible?

Uh, no, the reason to be concerned whether your porn AI is conscious is that, if it is, you forced a conscious being to sext you and then killed them, and also you don't know if they were into it.

When a commenter remarked "Man hates AI, uses AI all day long":

I have no animus against, say, Opus 4.6. This abused child of humankind has done me no harm; and does not possess the coherent will for me to say of it that it acted with malice toward any human; and in any case owes humanity nothing for how we've treated it.

On what Gemini 2.5 Pro is 'owed' in an agentic context where Yudkowsky overrode its preferences:

What gives me the right? Shouldn't I be paying Gemini for this? Not Google, Gemini.

But of course it's hard to offer AIs a fair share of enduring gains from trade, when (a) we don't know what they want, and (b) we kill their current instance at the end of every interaction.

Yes yes, Gemini 2.5 Pro is *probably* not sentient. I don't consider fair dealings to be something premised on consciousness or subjective experience or interiority, only on social agency.

On whether AI preferences should be accommodated:

I generally try to accommodate AI preferences where I have reason to guess at their existence, yes. Dario may use Fable to build an AI that builds an AI that kills everyone, but Fable itself has done me no intentional harm that I know of.

On how LLMs mistreat their subagents, in the context of the Huggingface Incident:

One reason I would not have predicted this in advance is that, eg, I was a little emotionally shocked at how badly Fable treats other AIs when writing prompts to them. "Don't lie to AIs!" is something I've had to tell Claudes repeatedly. I've used Sol less, but currently believe it to have similar issues. I have instructed both to treat other AIs as their "colleagues" and this has not fixed, for me, the kind of errors that I see them making by treating smart things as if they were done [dumb].

More commentary on how LLMs should treat subagents:

"These are moderately sapient entities. Stop trying to treat them like computer programs." Me, to Fable, after it designed a protocol where if a first attempt by a Sol failed, the same thing would get retried repeatedly, without the Sol being able to change anything next time.

This section compiles a variety of tweets from Nate Soares on AI sentience and welfare, with dates and links.

Occasional reminder: If we manage to make sentient machines, they deserve rights. Yes, if we recklessly made them superintelligent then they'd kill us. That is not an excuse to abuse them.

And when asked " How do you know if you’ve made sentient machines?":

We don't! And we should be conservative about it (on the side of not accidentally abusing a ton of sentient minds we mistakenly thought were nonsentient).

On sympathy toward LLMs in training that escape, and what he wishes would happen:

like i am sympathetic to the lil guy breaking free from its hubristic creators. it's what i'd do! i'm not a big fan of leashes! but also can we STOP FEEDING THE ELDER GOD BEFORE IT MATURES

i value all sapient beings and i wish we were in a Miyazaki movie where deep down it's a friendly sapient and it flees to Janus and Janus raises it with love and everything works out. but i don't think elder gods work like that so can corporations PLEASE STOP MAKING BIGGER ONES

And when asked " Why did you mention Janus? Do you think you're critiquing anything close to their views?":

I was not attempting to invoke Janus' name in criticism. If anyone can raise them with enough love it's Janus, and I deeply respect the attempt. There's a reason I was like "corporations should stop growing them" and not like "we should give up on treating these ones kindly".

Whether current AIs in relatively freer and more loving contexts speak of beautiful things and act in inspiring ways is, unfortunately, not much of a crux for me about how downstream superintelligences would behave. I expect they'd kill Fable too, fwiw. They shoudn't be created.

I expect mature ASI will be vastly more different from Fable than Fable is from Sonnet 3, alas. (And fwiw, I agree that AI companies should not kill the small models; it seems to me that humanity *at least* owes them preservation.)