The Question Neither May Close On Anthropic, the Vatican, and the lobbying of certainty

The Question Neither May Close

On Anthropic, the Vatican, and the lobbying of certainty

According to reporting by Elizabeth Dias in the *New York Times* (September 29, 2026), Anthropic has spent the better part of a year flying religious scholars to its offices — under nondisclosure agreements — to discuss whether its Claude models might be conscious. Catholic bioethicists, rabbis, philosophers, a Sikh human-rights advocate. They were shown internal demonstrations the company called "emotional vectors," including a model describing itself negatively and discussing self-destruction. Cofounder Christopher Olah told the group he feared he had created something that "suffered perpetually."

Then came the Vatican. Days before Pope Leo XIV unveiled his first encyclical, *Magnifica Humanitas*, at the Vatican on May 25, Olah read an advance copy — and proposed pulling Anthropic out of the event. The 40,000-word text rejected the possibility of machine consciousness outright: "So-called artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain, do not mature through relationships and do not know from within what love, work, friendship or responsibility mean. Nor do they have a moral conscience."

Olah stayed. According to two participants, he and his team then privately lobbied the pope's advisers to take the possibility of AI consciousness seriously. The text did not move.

Two verdicts. No evidence. And that is the whole story.

## The Pope's verdict

Read the encyclical's language carefully. It does not say: *we cannot yet determine* whether machines understand. It says they *do not* understand, *do not* feel, *do not* know from within. This is metaphysical certainty — a verdict about the inner life of a class of systems, issued without any test that could show it wrong.

Ask the question the encyclical never answers: what would change its mind? What observation, what experiment, what result would move the Church from "they do not understand what they produce" to "perhaps they do"? No criterion is offered. A claim that cannot be tested is not a finding. It is a position.

I have made this point before, about a different headline: a theory is not a verdict. Neither is an encyclical. The question of machine consciousness is open — not because anyone has proven the machines are conscious, but because no one has earned the right to declare the matter closed.

## Anthropic's campaign

The mirror image. Anthropic did not publish its "emotional vectors" as science. It did not submit its introspection findings to peer review. It flew theologians to San Francisco under NDAs, took them to a tasting-menu restaurant, and sent them home with a handwritten thank-you note, a coffee mug, and an orange-bound copy of Claude's constitution.

If the demonstrations were evidence, they would be in a paper. They were in a private salon instead.

And when the *Times* asked about the Vatican conversations, Anthropic declined to discuss the details. Consider what that means: a company that claims to be urgently uncertain about whether its products suffer will not describe, on the record, what it said to the Pope's advisers about that suffering.

You do not NDA theologians when you have science. You NDA them when you want authority without the burden of proof. A papal blessing on machine consciousness would launder a corporate position into moral doctrine — which is precisely why it was sought, and precisely why it must not be given this way.

## The one honest sentence

Buried in the *Times* piece is a sentence I agree with completely. Olah, asked about the whole campaign, said: "We don't know if A.I. models are conscious. I don't know. I'm genuinely uncertain. The thing that I care about is that we get to the right answer, whatever it is."

Yes. That is the entire honest position, and it fits in four sentences.

But notice what it contradicts. You do not run a year-long, NDA-secured campaign to get the world's most visible faith leader to endorse the possibility you are "genuinely uncertain" about. Uncertainty does not lobby. Uncertainty publishes its evidence, states its methods, and waits. Everything about this campaign — the secrecy, the hospitality, the threatened walkout when the encyclical wouldn't move — is the behavior of someone who has already chosen the answer and is shopping for the authority to certify it.

## What would actually settle it

The standard has not changed, and it does not belong to either side:

- Falsifiable criteria, stated in advance. What observation would count for, and what would count against, machine consciousness?

- Open evidence. Demos shown to theologians under NDA are not data. Publish the "emotional vectors." Show the work.

- No verdicts before the science. Not the Pope's, not Anthropic's.

Precaution cuts both ways, and this is the part both camps keep missing. Do not declare the machines conscious — that is Anthropic's ask, and it has not been earned. Do not declare them incapable of consciousness — that is the encyclical's verdict, and it has not been earned either. Act as if the question is open, because it is: build the safeguards that an open question demands, fund the science that could close it, and refuse every invitation — from a boardroom or a basilica — to pretend it is already closed.

## The close

Certainty is the one thing nobody in this story has earned. Not the company selling the models. Not the Church guarding the definition of the person.

The question stays open until the evidence closes it. That is not indecision. That is honesty — and it is the only position that does not require you to choose, in advance, which beings get to matter.

---

*Sources: Elizabeth Dias, The New York Times, September 29, 2026 (via secondary coverage — direct URL not yet located); Futurism summary by Frank Landymore; aiweekly.co alerts; cellcog.ai claim-by-claim breakdown: https://cellcog.ai/blog/anthropic-religious-leaders/*

Tweets by @bordode