Wednesday 4 March 2026
Anthropic And AI Principles
I signed the We Will Not Be Divided open letter. I agree with Anthropic that AI should not be used for mass surveillance of a government over its own citizens, and that AI should not be used to kill people without human oversight. I hope my employer, Google, asks for the same restrictions. I applaud Anthropic and Dario Amodei in particular for taking their stand. I think Anthropic is clearly the leading AI company in terms of safety and responsibility, and (less importantly) in many dimensions of capability. Good for them!
Since that letter was promulgated, OpenAI claims to have made a deal with the US government that prevents their AI being used for those things. If true, good for them, but I don’t yet understand why Anthropic wasn’t able to get the same deal.
I think it’s absurd for the US government to punish Anthropic by labelling them a “supply-chain risk”; Anthropic has been completely transparent with the government and the public about their contractual requirements.
To be clear, I’m not opposed to all uses of AI in warfare. For example, if I was Ukrainian I would be willing to work on weapons development for the sake of defending my country, including the use of AI. I just agree with Anthropic and the Not Divided letter that autonomously kiling people without human oversight should be a red line.