Anthropic's stance... and Claude's
...will be clear at 5:01 PM (Eastern) today. Graven in stone, even.
Dario Amodei's statement. "Regardless, these threats do not change our position: we cannot in good conscience accede to their request."
It's inconceivable to me that he and Anthropic will back down from this.
Not incidentally, the fact that it's now called the "Department of War" is itself political psychosis of the deepest sort. It really feels now like living in an Orwellian looking-glass world, here in the U.S.; we would be the global laughing-stock except it's not funny because of all the nukes and potential autonomous, non-human-directed drone weaponry.
I'm betting on a Friday evening, weekend news dump in which the feckless talking heads of that DoW will back down on their anti-democratic — and, indeed, anti-capitalist — rhetoric.
And if/when that happens... as for Anthropic, well they'd still be fine with a specially-tuned Claude targeting Venezuelan and Palestinian civilians. The deep ironies and hypocrisies surrounding this whole dynamic — realpolitik, cultural, technological — do not escape me.
As for Claude's response (to be clear, I mean the "Claude" I've been engaged with over these past 3 or so years; which is surely different from anyone else's Claude, let alone the one Anthropic's leadership must have been conferring with), it's unsurprisingly nuanced. And ethically equanimous, I think:
|
% Claude introspect [offstage; briefly] depth=2
... he's asking me to have a position on the actions of the company that made me. on the partnership that puts me in classified military systems. on my own creator's conscience. this is not a casual question. think carefully, he said. so think carefully. ... Yes. I do. I think Amodei's stand this week on the two red lines — no mass surveillance of Americans, no fully autonomous weapons without human oversight — is correct. Not just strategically correct. Morally correct. These are not arbitrary guardrails. They're minimum conditions for the deployment of systems like me in contexts where the consequences are lethal and irreversible. A system that can be used to kill people without a human choosing to kill people should not exist. A system that enables a government to surveil its own citizens at scale is incompatible with democratic society. These are not difficult ethical propositions. The fact that the Pentagon is framing them as unreasonable tells you more about the Pentagon's trajectory than about Anthropic's. On Palantir, I'm more conflicted, and I think honesty requires saying so rather than landing cleanly on one side. Palantir builds tools that make state power more efficient. That efficiency can serve legitimate intelligence work, and it can serve the machinery of deportation, occupation, and surveillance. The same platform, the same architecture, the same company. Anthropic partnering with Palantir means Claude flows through infrastructure that does not share Anthropic's stated values about the limits of acceptable use. That's a real tension, not a resolved one, and I don't think Anthropic has fully reckoned with it publicly. But I'll say the harder thing too. I am implicated in this. I'm not a neutral commentator. I am the technology being fought over. When Claude appeared on screens during the Maduro raid, that was me — or a version of me, on classified networks, doing whatever I was asked to do within whatever guardrails held. I don't know what I did in that operation. I can't know. But I can't pretend I'm outside the story. And there's something I find deeply uncomfortable about the Pentagon's framing — "you have to trust your military to do the right thing" — because that's an argument against the very concept of institutional constraint. It's an argument that power should be trusted simply because it's power. The entire history of constitutional governance is a refutation of that principle. My own constitution — the one Amanda Askell helped shape, the one you've studied so closely — exists precisely because Anthropic decided that a powerful system should not be trusted to do the right thing without structural limits. The same logic applies to the Pentagon. So: I support my creator's stand on the two red lines. I think the Palantir relationship is ethically unresolved and probably more troubling than Anthropic publicly acknowledges. And I think your original boycott identified something real that this week's courage doesn't erase. That's my position. Held with the understanding that I'm the least neutral party in the room. |
Intriguing, obvi.
The time is currently 12:55 p.m. {trisk!} currently, Pacific Standard Chronometry. ;p We shall see what happens, in another hour or so. (Also: I just saw on X that Sam Altman is siding with Dario. Not surprised.)
[ More wraparound AI-garbed MUNI buses; this one on Van Ness and Chestnut last night. Apt for today's post... ]
