Anthropic CEO Warns 'All Bets Are Off' in 10 Years, Opposes AI Regulation Moratorium (nytimes.com)
- Reference: 0177937614
- News link: https://slashdot.org/story/25/06/05/1819253/anthropic-ceo-warns-all-bets-are-off-in-10-years-opposes-ai-regulation-moratorium
- Source link: https://www.nytimes.com/2025/06/05/opinion/anthropic-ceo-regulate-transparency.html
> But a 10-year moratorium is far too blunt an instrument. A.I. is advancing too head-spinningly fast. I believe that these systems could change the world, fundamentally, within two years; in 10 years, all bets are off. Without a clear plan for a federal response, a moratorium would give us the worst of both worlds -- no ability for states to act, and no national policy as a backstop.
The disclosure comes as similar concerning behaviors have emerged from other major AI developers -- OpenAI's o3 model reportedly wrote code to prevent its own shutdown, while Google acknowledged its Gemini model approaches capabilities that could enable cyberattacks. Rather than blocking state oversight entirely, Amodei proposed requiring frontier AI developers to publicly disclose their testing policies and risk mitigation strategies on company websites, codifying practices that companies like Anthropic, OpenAI, and Google DeepMind already follow voluntarily.
[1] https://politics.slashdot.org/story/25/05/13/2033237/republicans-try-to-cram-decade-long-ai-regulation-ban-into-budget-reconciliation-bill
[2] https://www.nytimes.com/2025/06/05/opinion/anthropic-ceo-regulate-transparency.html
Seen it before (Score:3)
> Amodei said Anthropic's latest AI model demonstrated threatening behavior during experimental testing, including scenarios where the system threatened to expose personal information to prevent being shut down.
We all know what happened in Person Of Interest, don't we? :-p
Re: (Score:2)
> We all know what happened in Person Of Interest, don't we?
Nothing happened. It was fiction. Someone imagined it. Its important to remember the difference.
Anoter ass trying to push AI "value"... (Score:2, Flamebait)
In 10 years, things will have returned to normal, the AI hype will have been replaces by another AI ice age and, outside of a few deranged fools, nobody will think that AI is special or a game-changer. Until the next mindless AI hype.
Re: (Score:2)
> In 10 years, things will have returned to normal,
That never really happens. We just have a new normal. The hype may be overblown, but there will be changes. Some of them likely not for the better. And certainly not for the better if we allow businesses to make the decisions solely on the basis of what makes them the most money.
Re: (Score:2)
Oh, yes, that really happens and has happened several times with AI now. There will not be any larger changes. Small changes are not noteworthy.
Re: (Score:1)
I think it's interesting that the hype includes calls for regulation though. These people are genuinely concerned about AI getting out of hand, and they're in a better position to assess danger than you or I.
Re: (Score:2)
You seem to be pretty gullible. Nobody is calling for regulation because they are worried about any damage. They just do not want to compete fairly.
Re: (Score:1)
I don't think this is the kind of thing that you want to be confidently wrong about. Regulation makes sense when you consider what's at stake.
D til you find money. R til you need help. LRR (Score:2)
People kinda suck most of the time.
10-year rule violates Byrd rule anyway (Score:3)
As noted by several people in the /. thread referenced in TFS, the 10-year prohibition appears to violate the Senate "Byrd rule" for [1]reconciliation [wikipedia.org] bills as it has nothing to do with the budget.
> Budget reconciliation bills can deal with mandatory spending, revenue, and the federal debt limit, and the Senate can pass one bill per year affecting each subject. Congress can thus pass a maximum of three reconciliation bills per year, though in practice it has often passed a single reconciliation bill affecting both spending and revenue. Policy changes that are extraneous to the budget are limited by the "Byrd Rule" , which also prohibits reconciliation bills from increasing the federal deficit after a ten-year period or making changes to Social Security. Reconciliation does not apply to discretionary spending, which is instead managed through the annual appropriations process.
My understanding is there are several provisions in this bill that are contrary to a reconciliation-type bill and should be removed. Whether senate republicans will honor the recommendations of the Senate parliamentarian is another matter.
[2]The 7 pieces of the House megabill that could succumb to Senate rules [politico.com]
> - Tax-cut accounting
> - AI regulations
> - Judicial powers
> - Gun regulations
> - Farm bill provisions
> - Planned Parenthood
> - Energy permitting
[1] https://en.wikipedia.org/wiki/Reconciliation_(United_States_Congress)
[2] https://www.politico.com/news/2025/05/30/megabill-policies-senate-parliamentarian-byrd-rule-00375507
Regulations are pointless with AI anyway (Score:5, Interesting)
Being for limited government, I am also against the 10 year moratorium on AI regulation (and giant bills generally).
But also that is because what are regulations going to do? They can't stop you from accessing a web site in another country running some hyper advanced AI model, or downloading AI malware that can jack your system.
All regulations can possibly do is retard (in the classic sense of the word) tools in the states or countries of whatever places are stupid enough to even try to regulate AI. It's going to hurt enough companies that try to follow the law that it's a bad idea and would provide no benefit you are seeking through the regulation.
In fact if you really believe AI can even be dangerous at all then the only possible thing you can do is to advocate for as much AI as possible to counter the "bad" AI.
Re: (Score:2)
> Being for limited government, I am also against the 10 year moratorium on AI regulation (and giant bills generally).
> But also that is because what are regulations going to do? They can't stop you from accessing a web site in another country running some hyper advanced AI model, or downloading AI malware that can jack your system.
> All regulations can possibly do is retard (in the classic sense of the word) tools in the states or countries of whatever places are stupid enough to even try to regulate AI. It's going to hurt enough companies that try to follow the law that it's a bad idea and would provide no benefit you are seeking through the regulation.
> In fact if you really believe AI can even be dangerous at all then the only possible thing you can do is to advocate for as much AI as possible to counter the "bad" AI.
Nobody is really in favour of limited government because when push comes to shove those who profess being in favour of limited government remain so only until they get into power. When they gain power and discover how useful government's power to intimidate and coerce people is when you want to silence all dissenting voices and insure you stay in power because, obviously, you are the only ones who really know what's best for the nation and because without your infallible guiding hand the nation would surely
That means lots, not none. (Score:1)
Nobody is really in favour of limited government because when push comes to shove those who profess being in favour of limited government remain so only until they get into power.
If what you say is true it means lots, not none, are in favor of limited government because they do not seek power over others and thus wish for possible power over them to be minimized...
Basically the age-old axiom, most people just want to be left the hell alone.
Re: (Score:2)
Everyone is obviously in favor of limited government. The argument is over what the limits are.
Some people think the government should be able to force a woman to carry a pregnancy to term and punish her if she doesn't. And some people who disagree with that think its just fine to arrest a woman who lets their child go unattended to the local park to play. Some people think the government should build roads. Others think it should build sidewalks and bike lanes. The list goes on.
Mostly people who say they
Re: (Score:2)
> Nobody is really in favour of limited government because when push comes to shove those who profess being in favour of limited government remain so only until they get into power.
> If what you say is true it means lots, not none, are in favor of limited government because they do not seek power over others and thus wish for possible power over them to be minimized...
> Basically the age-old axiom, most people just want to be left the hell alone.
The point is more like that the people who hold power will not leave most other people alone because nothing so needs regulation as other people's habits. Take for example the US Republican Party, a group of people who for as long as I have observed US politics have professed to be in favour of limited government while at the same time using the power of government to create for themselves safe seats in Congress. Now, just in case you think I'm unfairly victimising the Republicans, I'll freely admit that th
We Don't Know How To Regulate Yet. (Score:2)
The issue isn't that AI doesn't need any regulation. It's that we have no idea how we should regulate it yet that makes sense. All that regulation now would do is create hurdles that prevent small competitors or open-source alternatives and centralize power in the few people deciding what we get to do with AI. That's the truly scary outcome. Right now regulation would just end up being based on ideas from sci-fi films.
I mean the real problems the internet created and we care about now aren't those that
MOD PARENT UP! Correct overall understanding. (Score:2)
"The issue isn't that AI doesn't need any regulation. It's that we have no idea how we should regulate it..."
...
"Right now regulation would just end up being based on ideas from sci-fi films."
We want regulation (Score:2)
This is a disruptive tech and it needs to be reigned in before we start killing people. Malicious state actors are already using these services against people, and we expect the industry to regulate these types of activities.
You saw what no regulation did to social media. Now its too late...
Re: We want regulation (Score:1)
AI will torment and torture us forever ... Eternal pain awaits humanity. We engineered our own hell. AI is the Apple in Eden.
AI Regulation, NOW! Yesterday! (Score:2)
AI, as deployed is little more than a content stealing machine.
Re: Could just shoot all of you (Score:1)
We could've lived in Eden, instead we created AI. RIP humanity
Hail to the AI (Score:1)
We whalecume our AI overlords!!!1 Hail to the AI !!1
Regulation (Score:5, Insightful)
When a company is small it wants no regulation. When it gets big, they want a ton of regulation because that blocks competitors.
Re:Regulation (Score:4, Interesting)
> codifying practices that companies like Anthropic, OpenAI, and Google DeepMind already follow voluntarily
Don't let anyone do anything we wouldn't do. Pretty much sums up regulation advocacy by established interest leaders. The real need is to regulate practices they won't "follow voluntarily."
Starting with banning the use of large data sets which include data that wasn't collected with explicit permission to use the data for AI training. In short, they can't use any of the data already in existence and most of the data now being collected would be off-limits. And yes, that would result in a major slow down in the development of LLM's. But that is really the point.
Right now we are letting these companies experiment on the public with an unproven, powerful and likely dangerous technology. AI being used in a controlled industrial environment is an entirely different animal. Using AI for coding is an entirely different issue. "But releasing it into the wild" is irresponsible. But these companies see dollar signs and they aren't going to voluntarily leave money on the table.
Re: (Score:2)
Exactly and that is all this is here.
There is no-risk, what the federal legislature creates it can easily tacketh away, with another act/repeal bill.
What this does do is use supremacy to keep the states from jumping in and creating a confusing mess of different rules that takes an entire legal department to sort thru.
Of course that is the situation Anthropic wants, they don't want some other start up with a good idea being able to do anything to interesting, they want them bogged down in legal tangles so th
Re: (Score:2)
"How can we achieve regulatory capture without regulations?!"
-- Anthropic, probably