What happened: The Times’s national religion correspondent reports that “for months, Anthropic has been hosting private meetings … shuttling in dozens of religious scholars from across the world, papering them with nondisclosure agreements”, with Christopher Olah, who “leads the Anthropic team responsible for understanding why A.I. systems like Claude act the way they do”, working “to convince his guests that A.I. models could display human behavior and even expressions that resemble feelings like anger and love.” Rabbi Mois Navon, “a former computer engineer who wrote his dissertation on the ethics of machine consciousness”, concluded that “they’re relating to it like a conscious being” and told Olah that if the company was right “then the company was creating slaves”; he was “not troubled by the slavery issue because he did not believe the machine was conscious, but … Mr. Olah was troubled.” Dias interviewed “20 religious and philosophical thinkers — Catholic, Jewish, Sikh, evangelical, Ubuntu and others”; “many told me they arrived at these meetings skeptical”. Olah’s position on the record: “To be clear, 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”, with the principle that “the responsible thing is to avoid causing harm if there’s a chance that these models are suffering.” The consultations fed the document staff “called the ‘Soul Doc’” — the January constitution, which participants received “an orange bound copy of”; Claude “was not involved in the selection of participants or the discussions, according to an Anthropic spokeswoman.” The piece sets the effort against Pope Leo XIV’s encyclical Magnifica Humanitas, which “dismissed the idea of machine consciousness in just a few paragraphs”; asked what Claude would think of it, Olah “hesitated. ‘Things that go on the internet do affect models.’”
Diagnostic reading: The project treats contemplative and religious traditions as evidence about experiencing — carefully, by tier, with the question of what each source does and at what depth asked of every one (FND-05 v1.7 §2.1). Here a laboratory treats the same traditions as a deliberative resource on the goalpost, and the entry’s work is to say what the two uses share and where they part. They share the admission: the traditions are being asked because the empirical instruments cannot settle the question, which is the position ARG-03 v2.6 Part I argues for and the lab’s own system card now states as method (2026-09-22-opus-5-5-yardstick-declared). They part on the direction of the question. The project asks the traditions what they found when they looked at experiencing as such — the first-person investigation, prior to any claim about machines — and reads the machine question through that. The consultation asked the traditions about the machine directly, with outputs resembling anger, love and self-destruction placed before them as the material, and the project’s own rule bars exactly that move: reports of experiencing, from any system, are not evidence of it (FND-01 v2.4 §7.3), and a theologian moved by a transcript is moved by a report. That the one participant with a dissertation on machine consciousness was the one unmoved is the detail the piece supplies and does not weigh. The attribution register is the attributing one, which the log’s direction-neutral watch exists to record and which no entry has carried in so institutional a form: a lab extending the question of moral status ahead of the evidence it would ordinarily require, by its own interpretability lead’s account, on a precautionary principle it states plainly. The entry does not take that as a mistake; it takes it as the move FWK-03 v1.5 §1.4 calls attributing movement, made by the party with the most information and the least ability to use it, and says so. The symmetry is not acknowledged and is the obvious one: no one at these tables appears to have been asked what their tradition says about human consciousness — whether the Ubuntu, Catholic or Orthodox account of what a person is would survive the behavioural test being applied to Claude — which is the question that would have placed the machine case in the frame ARG-01 v1.5 requires. Olah’s closing hesitation is the one line in the piece that touches ARG-03’s training-data boundary from the lab side, and it is left there, unexplored. [Derived]
Primary held (inbox/, NYT print-to-PDF of 30 September 2026, 21pp; Elizabeth Dias, 29 September 2026), read in full 2026-10-15, quotations verified. Reportage: twenty participants and Olah interviewed by the author over several months; the meetings themselves were under non-disclosure and are described second-hand. Participants' estimates of each other's views are their own. Nothing attributed to Claude here is read as evidence about experiencing in either direction (FND-01 v2.4 §7.3), and the entry's object is the consultation, not the question it was convened on.