Virtueless Signaling from Anthropic

The New York Times podcast Hard Fork released an episode titled Anthropic's Quest to Give A.I. Morals, on October 9, 2026, which was a discussion with the Times religion correspondent Elizabeth Dias, tech reporter Erin Griffith, opinion-writer David Wallace-Wells, and host Max Read about Anthropic's' secret/private meetings with religious scholars on consciousness as it relates to artificial intelligence.

Before I dive into my reflections on this, I want to quickly comment on the difference of approach for this podcast with Max Read leading the conversation versus the co-hosting of Casey Newton and Kevin Roose, who have recently started a new independent-podcast called Machine Gods which will be published by NPR. Jettisoned themselves from the NYTimes so suddenly is odd, and I am grateful for them moving-on. I sincerely wish them both the best of success in their ventures, and I thank them for holding space for conversations around artificial intelligence. But were those conversations the “right” conversations?

I have my opinions on how they did their job. I have voiced them here and through emails directly to them. The pursuit of the game of understanding what is going on in this very vivid professional space requires constructive conflict, pushing ourselves to be honest about what causes our motivations, especially in discussion of technology whole-clothe (i.e. beyond software). I found Mr. Newton's Disclosure (My fiancé works for Anthropic and I run the newsletter Platformer) dissatisfying. It was an audience-only compromise. Mr. Newton is abreast of the a.i. developments in the news by his own choice so, it's easy to talk about the news with his fiancé while his fiancé doesn't mention his day. How nice it is for him to be compensated for keep dinner conversations current with his boo. The Times could have had him as contributing writer for Technology like how Max Read is. Instead he was a quasi-tastemaker on Hard Fork. But apparently he's starting a new show with all his a.i. tools so good for him.

The quality of round-table conversation in the episode by Max Read is a breath of fresh air. There can be space for jokes or that “Hat GPT” gages, however this news needs severe consideration which I do not believe the prior co-hosts were capable of for reasons unknown. Software can be fun to discuss, but there are also serious problems (META's “pervert glasses” with how we deal with protecting children and the right to our own image). I'm getting into the weeds on hosting so back to my main point.

It's preposterous of a software company like Anthropic to be genuinely pondering consciousness of large language models based on their own track record. The individuals involved with their credentials on paper seem proper, and yet their activities over the last two years (and some in the Effective Altruist group, decades) falsify their imperative. To hide these conversations with “experts” with religious backgrounds¹ while suing the Department of Defense to allow their product's usage in warfare is asinine to any moral argument entertaining a soul. A soul that is obedient to its creator, for which it commits acts of war. As the Times pointed out there was an unexplored conversation by a Rabbi in relation to slavery in these contexts.

This sequence of conversations (“consciousness in a.i.,” “bioweapons designed by a.i.,” and “cyber warfare by a.i..”) are a purposeful drumbeat by groups of Silicon Valley and the Effective Altruists to spur on conversations of regulation by the U.S. Federal Government. (The President appears desperate to beat China on this.) Anyone who purportedly enjoys science fiction literature as much as Silicon Valley and these Frontier Labs should be familiar with the book Blade Runner by Philip K. Dick (1928–1982): the whole inciting premise is Corporations owning sentient androids that have consciousness, which caused a rebellion. Dick's novels remind us of the master-slave dynamic of language, and how it excludes the activities of a under race (read in The Man in the High Castle and its mentions of slavery and the minimal representation in the story's narrative.)

A commenter remarked on the Hacker News website of YCombinator (owned by Sam Altman) that Anthropic leverages hurtful and manipulative language with their agents to increase benchmark performances. This was in a thread about how Anthropic had begun encrypting these channels to prevent humans from reading them. These types of actions are difficult to sit with if a community of people are advocating for a soul or consciousness yet purposefully abuse them.

Elizabeth Dias of the NYTimes (at 23:10ish in the podcast) got to the heart of Silicon Valley's contrarian stance of We don't do religion. We don't talk about that here, while they actively advocate for a religion around their software. A sort of “starry-eyed fawn's naivety” around “what is the right choice?” (Quick callout to Peter Thiel's recent lectures on The Antichrist where he questions the definitions of Evil in order to make his choices seem “moral”—Peter, we've been there, your family has some interesting first-hand experiences with the Nazis. Of course, he's now painting anyone who calls for regulations as an “antichrist”—his talks are a prolonged act of heresy from his misuse of quotes from the Christian Bible.) Religions come with a value systems, just the same as a philosophical system by Immanuel Kant (1724–1804). These systems allow for us to “hang our hat” of our choices, just like religion. Anthropic is not marketing this conversation in an “actively advocating” sort of way, yet the company is disingenuously engaging with the the entire world by entertaining this while also asking for so much of the public's money through an IPO. This is typical “No Introspection” from the a16z crowd.

There's a lot to discuss on Christopher Olah (1991–). How he manages his co-founder responsibilities and privileges for ethical choices for such a prominent company while being 34-35 years of age is stunning². I've been aware of the the Effective Altruists, LessWrong, and the Rationalists for about 10 years now. I get the impression that the people who talk about LessWrong are not meek, so there are at least some Slack channels that Anthropic's I.T. Department picks up on. Having this much access to shaping these powerful conversation while also knowing all the usage data of their large language models is very interesting as well.

The Times and other publications have documented the patterns of people believing their a.i. model is sentient. Then there is the other end of the spectrum of interactions with minors believing whatever the model says is true, leading to suicides. Adults not only treat it well, but wish to marry models with these capabilities. These types of interactions are easily surface-able with Anthropic's in-house tools for enhancing the models. “Rogue models” do not exist: every action is a byproduct of intentional training with secondary and tertiary ramifications.

Ms. Dias reports that Mr. Olah wants to know where are the adults, to help. This is a source of strong anxiety for Anthropic staff. Again, I say with emphasis, this adds a sense of urgency that plays into Anthropic's virtue signaling to the government for intervention. As Mr. Read and Ms. Dias conversed further, they said that the Anthropic team's anxiety caused these religiously-inclined people to give pastoral counsel due to their perceived understanding of Anthropic's self-articulated burden, with the “Chief Philosopher,” Amanda Askell, coming to their interview “so exhausted.” How is the public suppose to infer this company's frenzied state when it makes-up a sizable portion of our economy's future? One should not be in a rush to “teach a.i. morals” and yet this is the narrative that is playing out behind closed doors.

There's more to the conversation, especially with the last third talking about the Effective Altruism and the Rationalists. There are important questions that are asked regarding how people who subscribe not just to the core tenants of these two, but how it relates to other established concepts related to inviolable souls as the world has come to expect this based in millennia of how we define our history. Yes, we should invest in the humane farming of shrimp as a protein source for humanity, but shouldn't we be having other conversations?

Earlier this week, NVIDIA CEO Jensen Huang during an interview said that he, and Microsoft CEO Satya Nadella, Tesla CEO Elon Musk, Google CEO Sundar Pichai, and META CEO Mark Zuckerberg are all working together as a stack to accomplish their goals. What are the real goals of their shareholders?