> In their blog post, Anthropic defended its decision by saying the jailbreak isn’t serious. That is not what the trusted partner and the USG believe; nor is that kind of minimizing language consistent with Anthropic’s brand as the AI safety company.
This is what makes me feel Sacks is speaking the truth here, despite my generally not trusting him (due to MAGA sycophancy) or this administration in general. Given Anthropic and Dario in particular are so alarmist about safety, even a small jailbreak should cause them to pull back and fix it first, right? Didn’t they say Mythos is very dangerous in the wrong hands? How can you take any chances if that’s the case? It’s just not consistent to minimize things - and I feel they probably didn’t want to admit to the world that their own safeguards aren’t good enough either, because that would harm their business if they had to pull back their models and stick to their safety views. But if they didn’t pull these models, they would be admitting that safety is theater for regulatory capture and that it doesn’t really matter.
By minimizing the jailbreak, they’re trying to have it both ways. And that feels dishonest. But also any mature executive would know how this would look. Even if they were correct about it, Anthropic should know how all of this looks to the outside world. The fact that Dario doesn’t, shows they shouldn’t be the ones in charge of this capability. Both OpenAI and Anthropic need brand new professional leadership.
As for Amazon - they aren’t going to raise alarms randomly. AWS is very trustworthy and well run. I believe their security researchers genuinely were worried about the impact of the general public having Mythos access effectively by jailbreaks on Fable. Plus they might be the largest owner of Anthropic equity outside of Anthropic itself.
"even a small jailbreak should cause them to pull back and fix it first, right"
You do realize that LLMs are summarizations of vast numbers of weights, don't you? You don't "fix" a weight and suddenly everything is alright. You can only probe constantly in a vast space and see if the results you can command matter or not.
From what I understand Amazon's lawyer was telling the CEO that it's better to report a vulnerability of a potential cyber weapon (jailbreak) to protect Amazon from liability.
And then US chamber of commerce protected itself from liability as well to take export control law seriously after Antropic's CEO categorized Mythos without the cyber query filter as potential cyber weapon.
The problem here is that there's no real technical way to protect against the queries, just making them more expensive to create, as the models are getting smarter.
Also as the main jailbreak ,,technique'' is splitting the task into subtasks, and the main moat of Mythos is that it can solve more complex tasks, I also wouldn't categorize this jailbreak as serious.
Remember when the admin tried to kill anthropic with a ridiculous national defense supply chain order that the courts blocked - I'm not sure why anyone would believe them credible now
David Sacks is probably the least credible guy associated with Trump administration AI policy. Yes, worse than Hegseth and Trump himself. Consistently duplicitous.
On the other hand, Anthropic really needs to up its game here. I am sympathetic to their position, but their actions betray a lack of understanding of the seriousness of the situation. If we're to believe that only they can align a super intelligence, they should start by aligning the Trump administration into a productive relationship.
Everyone I know who works at Anthropic genuinely believes in safety and has deep concerns about the potential existential impact of AI; the "they are just hucksters pretending to care about it so they can run a scam" point of view isn't at all close to reality.
But, granting it for the sake of argument, yeah, they're clearly failing there: if other companies are able to outcompete them on the regulatory capture front, then if that's their strategy then they're losing.
Hegseth would try to nationalize them. He wouldn’t succeed. But he’s an idiot who trashes things without any clear benefit to himself, and the ensuing fight would probably push Anthropic’s IPO window out of 2026.
It's rather obvious to me once the Chinese and Russian hackers get a hold of this it will be "jailbroken" overnight, and all mayhem and anarchy will break loose.
Unfortunately we cannot trust this guy. He’s deep in the Trump administration and has been doing everything in his power to try and pick other companies to win over Anthropic. Just before all this happened, he was going on a rant about how Anthropic was doing too much to disable their models for safety / regulatory capture reasons!
There was a time I used to listen to the All In Podcast. Then Trump became president and these guys started gloating with glee about all the breakage that Musk and Trump were causing. Was quite the shock to me to see the all out happiness. That's when I stopped listening to Sacks and his crowd.
Ignore their words entirely and pay attention only to the real world effects of their actions, whilst concentrating on your own plans and adjusting accordingly.
In the end, core infrastructure projects need to get their shit together, use Mythos to patch their software, then the tool needs to be available to the public. There is no way to keep the lid on advanced coding AIs.
I do not believe it when an admin ally says the government is acting in good faith or that there is no bitterness or history with DoD (not DoW, another indicator he's full of it).
The administration has made it official policy to fire apolitical competent federal workers. Don't tell me they have a good heart and just want things to be done safely.
Dario, buy a few million of Trumpcoin and you'll be fine.
But the world cannot use Mythos. Only some “trusted partners” of Anthropic can. Whatever that means.
The rest are stuck with Fable, which will refuse to help find vulnerabilities in your own software.
> In their blog post, Anthropic defended its decision by saying the jailbreak isn’t serious. That is not what the trusted partner and the USG believe; nor is that kind of minimizing language consistent with Anthropic’s brand as the AI safety company.
This is what makes me feel Sacks is speaking the truth here, despite my generally not trusting him (due to MAGA sycophancy) or this administration in general. Given Anthropic and Dario in particular are so alarmist about safety, even a small jailbreak should cause them to pull back and fix it first, right? Didn’t they say Mythos is very dangerous in the wrong hands? How can you take any chances if that’s the case? It’s just not consistent to minimize things - and I feel they probably didn’t want to admit to the world that their own safeguards aren’t good enough either, because that would harm their business if they had to pull back their models and stick to their safety views. But if they didn’t pull these models, they would be admitting that safety is theater for regulatory capture and that it doesn’t really matter.
By minimizing the jailbreak, they’re trying to have it both ways. And that feels dishonest. But also any mature executive would know how this would look. Even if they were correct about it, Anthropic should know how all of this looks to the outside world. The fact that Dario doesn’t, shows they shouldn’t be the ones in charge of this capability. Both OpenAI and Anthropic need brand new professional leadership.
As for Amazon - they aren’t going to raise alarms randomly. AWS is very trustworthy and well run. I believe their security researchers genuinely were worried about the impact of the general public having Mythos access effectively by jailbreaks on Fable. Plus they might be the largest owner of Anthropic equity outside of Anthropic itself.
"even a small jailbreak should cause them to pull back and fix it first, right"
You do realize that LLMs are summarizations of vast numbers of weights, don't you? You don't "fix" a weight and suddenly everything is alright. You can only probe constantly in a vast space and see if the results you can command matter or not.
Does anyone still believe anything this guy says?
When this tool insists it has nothing to do with their vendetta against Amodei, you know that's exactly what this is about.
From what I understand Amazon's lawyer was telling the CEO that it's better to report a vulnerability of a potential cyber weapon (jailbreak) to protect Amazon from liability.
And then US chamber of commerce protected itself from liability as well to take export control law seriously after Antropic's CEO categorized Mythos without the cyber query filter as potential cyber weapon.
The problem here is that there's no real technical way to protect against the queries, just making them more expensive to create, as the models are getting smarter.
Also as the main jailbreak ,,technique'' is splitting the task into subtasks, and the main moat of Mythos is that it can solve more complex tasks, I also wouldn't categorize this jailbreak as serious.
Remember when the admin tried to kill anthropic with a ridiculous national defense supply chain order that the courts blocked - I'm not sure why anyone would believe them credible now
https://xcancel.com/DavidSacks/status/2065853007619588171
David Sacks is probably the least credible guy associated with Trump administration AI policy. Yes, worse than Hegseth and Trump himself. Consistently duplicitous.
On the other hand, Anthropic really needs to up its game here. I am sympathetic to their position, but their actions betray a lack of understanding of the seriousness of the situation. If we're to believe that only they can align a super intelligence, they should start by aligning the Trump administration into a productive relationship.
Alternatively, they could stop advocating for government control of “dangerous” (aka competitively advanced) models.
If you are running a pull-up-the-ladder play, it helps to be ensconced above it first.
Everyone I know who works at Anthropic genuinely believes in safety and has deep concerns about the potential existential impact of AI; the "they are just hucksters pretending to care about it so they can run a scam" point of view isn't at all close to reality.
But, granting it for the sake of argument, yeah, they're clearly failing there: if other companies are able to outcompete them on the regulatory capture front, then if that's their strategy then they're losing.
might be easier to align the super intelligence than Trump
I have a question. If A were to basically just release Mythos as-is, what would happen? What outcome are they so afraid of?
They're afraid that it will find vulnerabilities in everything and create thousands of new zero day exploits, apparently
> What outcome are they so afraid of?
Competitor distillation
Hegseth would try to nationalize them. He wouldn’t succeed. But he’s an idiot who trashes things without any clear benefit to himself, and the ensuing fight would probably push Anthropic’s IPO window out of 2026.
Cyber-attacks to start, and real-world terrorist attacks/bombings (inc. chemical/biological weapons) later.
It's rather obvious to me once the Chinese and Russian hackers get a hold of this it will be "jailbroken" overnight, and all mayhem and anarchy will break loose.
Unfortunately we cannot trust this guy. He’s deep in the Trump administration and has been doing everything in his power to try and pick other companies to win over Anthropic. Just before all this happened, he was going on a rant about how Anthropic was doing too much to disable their models for safety / regulatory capture reasons!
There was a time I used to listen to the All In Podcast. Then Trump became president and these guys started gloating with glee about all the breakage that Musk and Trump were causing. Was quite the shock to me to see the all out happiness. That's when I stopped listening to Sacks and his crowd.
The administration has a credibility issue, most certainly when it comes to rule of law and corruption.
Mythos being drummed up as a cyber weapon is not the same as it being a weapon -- and treating it as such is just dumb.
Yes, it may find vulnerabilities, but "find" is the keyword -- whatever it finds was already there.
What’s the game theory for understanding who to believe when they’re all fucking liars?
Ignore their words entirely and pay attention only to the real world effects of their actions, whilst concentrating on your own plans and adjusting accordingly.
In the end, core infrastructure projects need to get their shit together, use Mythos to patch their software, then the tool needs to be available to the public. There is no way to keep the lid on advanced coding AIs.
I do not believe it when an admin ally says the government is acting in good faith or that there is no bitterness or history with DoD (not DoW, another indicator he's full of it).
The administration has made it official policy to fire apolitical competent federal workers. Don't tell me they have a good heart and just want things to be done safely.
Dario, buy a few million of Trumpcoin and you'll be fine.
But the world cannot use Mythos. Only some “trusted partners” of Anthropic can. Whatever that means. The rest are stuck with Fable, which will refuse to help find vulnerabilities in your own software.
They need to get their shit together with anthropic and get access, then. I'm only mildly flippant.
Multiple vendors have probably run Linux through Mythos. I wonder if any patches are going upstream.
This makes much more sense. Whatever criticism for trump he's realist who does want the US to win the AI wars.