一个 Hacker News 帖子以“Did Anthropic ask for this?”为标题，截至发帖时获得 114 个点赞。
原文 · 未翻译
Did Anthropic Ask For This?
This past Friday, the US Government issued an export control directive that prohibits Anthropic from giving foreign nationals access to Claude Fable or Claude Mythos, their latest models.
I think Anthropic directly asked for this to happen.
My Case
A few days before this happened Dario Amodei, the CEO of Anthropic, published “Policy on the AI Exponential”:
The government should have the power to block or deter deployment of the model if it is determined, in light of third-party assessment, to present unacceptable risks. This power must be scoped to the above four specific risks and there must be protective measures against political favoritism or arbitrary decisions.
The government should have the power to block or deter deployment of the model if it is determined, in light of third-party assessment, to present unacceptable risks. This power must be scoped to the above four specific risks and there must be protective measures against political favoritism or arbitrary decisions.
Dario is known for writing about regulation and the direction of AI as an industry and Anthropic in particular, and what he says is taken very seriously and is considered a definitive statement of the company’s position. I take his statements this way, it is likely that the people who write our laws, and to whom he talks personally, take his statements this way, and it is likely that any judge he ends up in front of will also take his statements this way.
Since this is an official statement, let’s take it piece by piece to make absolutely sure he is asking for what happened to Anthropic two days later.
The government should have the power to block or deter deployment of the model
The government should have the power to block or deter deployment of the model
Yes. The government is blocking deployment of Anthropic’s model.
if it is determined, in light of third-party assessment, to present unacceptable risks.
if it is determined, in light of third-party assessment, to present unacceptable risks.
Yes. This assessment was made by Amazon, a frequent and serious government contractor which is generally trusted to handle high-security government, intelligence, and military contractor concerns.
This power must be scoped to the above four specific risks
This power must be scoped to the above four specific risks
We have to back up a bit to see what four risks these are, but the previous paragraph states:
cybersecurity, biological weapons, loss of control of AI systems, and automated R&D that could accelerate these other risks.
一个 Hacker News 帖子以“Did Anthropic ask for this?”为标题，截至发帖时获得 114 个点赞。
原文 · 保持原样，未翻译
Did Anthropic Ask For This?
This past Friday, the US Government issued an export control directive that prohibits Anthropic from giving foreign nationals access to Claude Fable or Claude Mythos, their latest models.
I think Anthropic directly asked for this to happen.
My Case
A few days before this happened Dario Amodei, the CEO of Anthropic, published “Policy on the AI Exponential”:
The government should have the power to block or deter deployment of the model if it is determined, in light of third-party assessment, to present unacceptable risks. This power must be scoped to the above four specific risks and there must be protective measures against political favoritism or arbitrary decisions.
The government should have the power to block or deter deployment of the model if it is determined, in light of third-party assessment, to present unacceptable risks. This power must be scoped to the above four specific risks and there must be protective measures against political favoritism or arbitrary decisions.
cybersecurity, biological weapons, loss of control of AI systems, and automated R&D that could accelerate these other risks.
Yes. Their third party report indicated that this formed a cybersecurity risk.
and there must be protective measures against political favoritism or arbitrary decisions.
and there must be protective measures against political favoritism or arbitrary decisions.
This is the only difficult part of the paragraph. Are there protective measures against political favoritism or arbitrary decisions?
I believe there are: they are called “courts”. Dario is as free as the rest of us are to file a lawsuit and go in front of a judge and tell the judge that he is the victim of political favoritism or an arbitrary decision. That is, in fact, one of the primary purposes of the legal system.
I do not see why I should believe that he meant something else. If he wanted there to be some “protective measure” other than the law, as it already exists, he has had ample opportunity for years to say so. Anthropic has never, to my knowledge, proposed that anyone working in AI should have new or different protections from the government than the ordinary course of the law. If Anthropic did believe there should be such a thing, and did not say so while pushing for stricter laws, it seems extremely negligent of them.
So yes. Dario asked for this. Anthropic asked for this. They got what they asked for, letter for letter and word for word.
Anthropic’s Position on Regulations
Not only did they ask for this, they asked for this for years, and they could not be talked out of asking for it. When people said that it might be bad to empower the government in this way, Anthropic was not sympathetic. Anthropic seemed to feel very strongly about this.
They have also, bizarrely, not changed their position even slightly as the government has changed hands and has mostly gone insane. They have continued to push for the government to have more power over AI, and they have continued to sell services to the government both directly and via intermediaries like Palantir and Amazon.
In my opinion, they mostly imagined these regulations applying to other people, especially open source projects, academics and smaller companies. Now that they are being subjected to the exact sort of regulation they have proposed, they do not like it.
I think all of this was extremely irresponsible of them, and I feel a good amount of schadenfreude that the leopard ate their face first.
Is This Politically Motivated or Arbitrary?
Probably at least somewhat. The Department of War appeared to try to kill Anthropic earlier this year. That was, in my opinion, an overtly political, ham-handed power grab over the company by the Secretary of War. So far the legal system appears to agree with me in that assessment.
What’s different about this?
First, that none of the clumsy political posturing or clear over-reach is involved. Based on reporting, this was negotiated by Bessent, Secretary of Treasury, and you will find no statements about Anthropic being “woke” in anything he has said about the matter because he has said nothing publicly about the matter at all. Export control directives are a clear power that the US Government actually has, unlike much of what the Department of War was asserting and demanding.
Second, and much more seriously, there is an unimpeachable record of public statements by Dario Amodei, CEO of Anthropic, and by Anthropic via official communications, that they believe Claude Mythos and Fable are of national security import because they pose serious bioweapon and cybersecurity risks, and that they should be regulated by the government. I am reasonably sure that a smart government lawyer would spend much of their time in court reading these statements into the record, if the government were sued over this.
Anthropic spends much of its time and energy talking about how dangerous AI is, and how it ought to be regulated as a military concern. They have been doing this for so long and so openly that it seems like people have gotten used to it, and no longer notice how weird and legally questionable it seems to be. What happens if the government takes everything Anthropic has said about itself both literally and seriously?
Dogs catch cars sometimes, and it’s often unfortunate when they do.
AI Companies Can’t Have It Both Ways
Broadly, AI companies like to talk up how cataclysmic and important what they’re doing is, and when the possible negatives are mentioned they tend to punt.
“The government” or “society” is meant to deal with all of those things. Well, now the government is — the actual government that really exists, and not an imagined one that only does good things and never does bad things.
I don’t particularly like this government action either. It does not seem likely at all that the US Government, as presently constituted, is capable of governing any radically new technology or an increasingly different future. If your plan is for the government to save us all later, you just don’t have a plan.
If no existing institution is able to govern the technology you’re making, you had better figure out making an institution that is or stop doing what you’re doing.
Nobody is going to care if you tell them that you thought the government or “society” was going to solve the problems you created.
Dario is known for writing about regulation and the direction of AI as an industry and Anthropic in particular, and what he says is taken very seriously and is considered a definitive statement of the company’s position. I take his statements this way, it is likely that the people who write our laws, and to whom he talks personally, take his statements this way, and it is likely that any judge he ends up in front of will also take his statements this way.
Since this is an official statement, let’s take it piece by piece to make absolutely sure he is asking for what happened to Anthropic two days later.
The government should have the power to block or deter deployment of the model
The government should have the power to block or deter deployment of the model
Yes. The government is blocking deployment of Anthropic’s model.
if it is determined, in light of third-party assessment, to present unacceptable risks.
if it is determined, in light of third-party assessment, to present unacceptable risks.
Yes. This assessment was made by Amazon, a frequent and serious government contractor which is generally trusted to handle high-security government, intelligence, and military contractor concerns.
This power must be scoped to the above four specific risks
This power must be scoped to the above four specific risks
We have to back up a bit to see what four risks these are, but the previous paragraph states:
cybersecurity, biological weapons, loss of control of AI systems, and automated R&D that could accelerate these other risks.
cybersecurity, biological weapons, loss of control of AI systems, and automated R&D that could accelerate these other risks.
Yes. Their third party report indicated that this formed a cybersecurity risk.
and there must be protective measures against political favoritism or arbitrary decisions.
and there must be protective measures against political favoritism or arbitrary decisions.
This is the only difficult part of the paragraph. Are there protective measures against political favoritism or arbitrary decisions?
I believe there are: they are called “courts”. Dario is as free as the rest of us are to file a lawsuit and go in front of a judge and tell the judge that he is the victim of political favoritism or an arbitrary decision. That is, in fact, one of the primary purposes of the legal system.
I do not see why I should believe that he meant something else. If he wanted there to be some “protective measure” other than the law, as it already exists, he has had ample opportunity for years to say so. Anthropic has never, to my knowledge, proposed that anyone working in AI should have new or different protections from the government than the ordinary course of the law. If Anthropic did believe there should be such a thing, and did not say so while pushing for stricter laws, it seems extremely negligent of them.
So yes. Dario asked for this. Anthropic asked for this. They got what they asked for, letter for letter and word for word.
Anthropic’s Position on Regulations
Not only did they ask for this, they asked for this for years, and they could not be talked out of asking for it. When people said that it might be bad to empower the government in this way, Anthropic was not sympathetic. Anthropic seemed to feel very strongly about this.
They have also, bizarrely, not changed their position even slightly as the government has changed hands and has mostly gone insane. They have continued to push for the government to have more power over AI, and they have continued to sell services to the government both directly and via intermediaries like Palantir and Amazon.
In my opinion, they mostly imagined these regulations applying to other people, especially open source projects, academics and smaller companies. Now that they are being subjected to the exact sort of regulation they have proposed, they do not like it.
I think all of this was extremely irresponsible of them, and I feel a good amount of schadenfreude that the leopard ate their face first.
Is This Politically Motivated or Arbitrary?
Probably at least somewhat. The Department of War appeared to try to kill Anthropic earlier this year. That was, in my opinion, an overtly political, ham-handed power grab over the company by the Secretary of War. So far the legal system appears to agree with me in that assessment.
What’s different about this?
First, that none of the clumsy political posturing or clear over-reach is involved. Based on reporting, this was negotiated by Bessent, Secretary of Treasury, and you will find no statements about Anthropic being “woke” in anything he has said about the matter because he has said nothing publicly about the matter at all. Export control directives are a clear power that the US Government actually has, unlike much of what the Department of War was asserting and demanding.
Second, and much more seriously, there is an unimpeachable record of public statements by Dario Amodei, CEO of Anthropic, and by Anthropic via official communications, that they believe Claude Mythos and Fable are of national security import because they pose serious bioweapon and cybersecurity risks, and that they should be regulated by the government. I am reasonably sure that a smart government lawyer would spend much of their time in court reading these statements into the record, if the government were sued over this.
Anthropic spends much of its time and energy talking about how dangerous AI is, and how it ought to be regulated as a military concern. They have been doing this for so long and so openly that it seems like people have gotten used to it, and no longer notice how weird and legally questionable it seems to be. What happens if the government takes everything Anthropic has said about itself both literally and seriously?
Dogs catch cars sometimes, and it’s often unfortunate when they do.
AI Companies Can’t Have It Both Ways
Broadly, AI companies like to talk up how cataclysmic and important what they’re doing is, and when the possible negatives are mentioned they tend to punt.
“The government” or “society” is meant to deal with all of those things. Well, now the government is — the actual government that really exists, and not an imagined one that only does good things and never does bad things.
I don’t particularly like this government action either. It does not seem likely at all that the US Government, as presently constituted, is capable of governing any radically new technology or an increasingly different future. If your plan is for the government to save us all later, you just don’t have a plan.
If no existing institution is able to govern the technology you’re making, you had better figure out making an institution that is or stop doing what you’re doing.
Nobody is going to care if you tell them that you thought the government or “society” was going to solve the problems you created.