Anthropic could have negotiated before USG publicly threatened to label them a supply chain risk. My guess is they were mainly limited by the erosion of their own morals and by Anthropic staff quitting, and they could have acquiesced with diplomatic language even within those limitations, maybe even after the crisis went public. Claude is only getting better, so the default path is building trust with the government. They could probably have found a better stand to take later, when they have more power.
Perhaps. I think writing things into contracts is a great way to make sure that they happen, and if the counterparty is unwilling to sign them into contracts, then this is a strong sign that you won’t be able to make it happen later. It would have significantly increased the adversarial relationship between Anthropic and the USG for them to politely remove it from the contract and then work hard internally to make sure that it never got used that way. Maybe it would’ve been worth it, but I’m not convinced.
Oh I don’t think they could have prevented USG from using Claude for mass domestic surveillance. Autonomous weapons maybe since it’s a reliability issue the military would agree with. They would need to sacrifice their principles in order to get Claude more integrated into the government, which could be good or bad for us but would have been in Anthropic’s interest.
I have a little stored thought which sometimes triggers, and it reads:
“If you find yourself being forced to choose between two or more extremely bad options that involve burning your values, your resources, or your life, the truth is that you lost around three moves ago and are living out the equivalent of a forced mate in chess. You’ve already lost, so stop playing and find a better game to spend time on if at all possible.”
Anthropic negotiated a great deal and gave up the practical limits that are relevant to the military about using their models for cyber attacks, censorship and disinformation campaigns in the process.
Anthropic could have negotiated before USG publicly threatened to label them a supply chain risk. My guess is they were mainly limited by the erosion of their own morals and by Anthropic staff quitting, and they could have acquiesced with diplomatic language even within those limitations, maybe even after the crisis went public. Claude is only getting better, so the default path is building trust with the government. They could probably have found a better stand to take later, when they have more power.
Perhaps. I think writing things into contracts is a great way to make sure that they happen, and if the counterparty is unwilling to sign them into contracts, then this is a strong sign that you won’t be able to make it happen later. It would have significantly increased the adversarial relationship between Anthropic and the USG for them to politely remove it from the contract and then work hard internally to make sure that it never got used that way. Maybe it would’ve been worth it, but I’m not convinced.
Oh I don’t think they could have prevented USG from using Claude for mass domestic surveillance. Autonomous weapons maybe since it’s a reliability issue the military would agree with. They would need to sacrifice their principles in order to get Claude more integrated into the government, which could be good or bad for us but would have been in Anthropic’s interest.
It’s a tough game to be in. Sometimes the only winning move is not to play.
I have a little stored thought which sometimes triggers, and it reads:
“If you find yourself being forced to choose between two or more extremely bad options that involve burning your values, your resources, or your life, the truth is that you lost around three moves ago and are living out the equivalent of a forced mate in chess. You’ve already lost, so stop playing and find a better game to spend time on if at all possible.”
Sadly, sometimes you don’t have the option of not playing.
Anthropic negotiated a great deal and gave up the practical limits that are relevant to the military about using their models for cyber attacks, censorship and disinformation campaigns in the process.