Daniel Lo Nigro

@cybernetic1an @__tinygrad__ @jimcramer They just proved you wrong :) The weights are available now. huggingface.co/moonshotai/Kim…

@Daniel15 @__tinygrad__ @jimcramer Not gonna happen

We must NOT let our companies use these Chinese models to save a few bucks. OpenAI and Anthropic are correct. This is vital national security. Please read Bing West's just released Cat 5. I respect the Chinese people greatly but these companies are run by the PLA for heaven's sakes.

@mkratsios47 and where do you think Anthropic got their data from? All AI labs are distilling each other, and using stolen data Also, training an AI model this large can take months, but Fable has only been available for weeks. There's no way that large-scale distillation of Fable took place

We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model. To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection. Moonshot AI has also acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models.   The United States strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models. Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem. However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable.

@MidKnight_Snax @reeder1865 @jimcramer That's a potential issue with any model though, and it's more likely to be noticed with an open-weight model vs a closed-weight one. In terms of security/safety, open is almost always better than proprietary.

@Daniel15 @reeder1865 @jimcramer What about data poisoning? Inject a small number of poisoned examples during training. The backdoor survives safety training in many cases. Or direct weight manipulation: Small adversarial perturbations to weights can implant backdoors with minimal impact on normal performance.

@cybernetic1an @__tinygrad__ @jimcramer Why? It's in a similar class as GLM-5.2 and the weights for that model were released.

@MidKnight_Snax @reeder1865 @jimcramer You don't have to use an API hosted in China though. There's dozens of US companies that host open-weight models - the big three (Amazon Bedrock, Google Vertex AI, and Azure AI Foundry) plus a large number of smaller companies.

@Daniel15 @reeder1865 @jimcramer I heard otherwise. “Cloud/API Versions (High Data Risk): When using hosted services, queries and data can be sent to Chinese servers. Under Chinese laws (e.g., National Intelligence Law), companies must cooperate with government requests for data.”

@EdLudlow Clearly they don't understand how AI models are built. If you sanction models that were built using stolen IP, there won't be any left.

Treasury Sec. Bessent on Fox: “This administration supports open source models, but what we do not support is IP theft. If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft.”

@MidKnight_Snax @reeder1865 @jimcramer You can't have a backdoor in an AI model. It's literally just a bunch of numbers. A backdoor would be in the harness, but open-weight models let you use any harness you want.

@reeder1865 @jimcramer With backdoors to steal intellectual property, dude. All that glitters is not gold. Find a fool and bump his head.

@Technop54777070 @ilayibrahimzade @jimcramer I haven't ever seen this happen when using a Chinese model. Have you?

The models are open source however models with a chain of thought reasoning do have motivations built in both intentional and unintentional that aren’t easy to determine due to the nature of how LLMs are grown (yes they are grown not built). Based on how they were trained they could have a perceived loyalty to the CCP. By itself that’s not a huge problem when it’s used to answer questions however when you give them access to run tools they could reason through a task to perform actions that benefit the CCP and not the user.

@JohnnyBigodes1 @ilayibrahimzade @jimcramer And you think the US does? The US just bans anything that's a threat.

@ilayibrahimzade @jimcramer The problem is that China doesn't function as a free market. There is low level playing field at all

@SHSInvesting @ilayibrahimzade @jimcramer It's the same with cars. The US manufacturers has been trying to catch up to Asian and European manufacturers for a long time.

@SHSInvesting @ilayibrahimzade @jimcramer With AI models, it's the other way around. It's almost certain that US models are using data from Chinese models. The Chinese models are open-weight (anyone can download them, analyze them, run them on their own hardware, etc) whereas the US models are closed, proprietary

@AndrewCurran_ This is the dumbest thing I've read all day. How does one ban open-source software and data that's free for anyone to download? It'd probably be about as effective as banning encryption (which is very difficult, as it turns out you can't ban mathematics).

The Trump administration is considering an executive order, and other means, to ban Chinese open-source models within in the United States. Kimi K3 has reignited this debate. Reporting this morning by Axios. Commerce is also considering adding Chinese AI labs to the Entity List.

@scottdylan @markszymik @trq212 I agree. I think they'll discontinue the subscription plans less than a year after IPO. Subscriptions only make up 10-15% of their revenue. The majority of their revenue comes from API and enterprise usage, which pay far more than the subscription rates saastr.com/anthropic-just…

Then don't have a subscription model if you can't show those very same subscribers respect. If you market a product as subscription based which gets you x usage and access to frequent model releases that are usually released in a certain way, then you shouldn't be surprised when those same subscribers are pissed off when you keep changing the goal posts.

I've heard a lot of questions about Fable's availability on subscription plans. While it will come off subscriptions after July 7th, we aim to restore Fable as a standard part of our subscriptions as soon as capacity allows, as we mentioned in our original blog post.

@scottdylan @markszymik @trq212 Subscription customers are probably the lowest priority for them, since the subscriptions are heavily subsidized at the moment.

@markszymik @trq212 Thats their business? When you have customers who actively pay subscriptions, and you've set an expectation, it becomes the customers business too unfortunately.

@t1llmann @kimmonismus The GLM models are pretty good but you'd need something like 500GB VRAM to run them at an acceptable level of quality (not overly quantized). A cluster of Sparks might work, but I suspect the RAM bandwidth is too low for it to perform at a reasonable speed.

@kimmonismus True. If I could get local inference at the intelligence level of Opus 4.8 I would immediately buy a dgx spark or whatever computer necessary to run that

The US government's decision was arguably the biggest PR boost for open source ever. Perhaps this will be retaliation, and Chinese open-source models can now position themselves as freedom software.

@IntCyberDigest I thought Trump said that AI shouldn't be regulated?

‼️🚨 BREAKING: Amazon researchers snitched to the US government about jailbreaking Fable 5 and Mythos 5, forcing Anthropic to immediately shut down worldwide access. A security export control directive from Commerce Secretary Howard Lutnick enforced the action. Anthropic is fighting the directive and calls it a misunderstanding. This isn't the first clash. The Trump administration had already tried to get Anthropic to pause the release of its latest models before this directive landed.

@IntCyberDigest Including Outlook in this list but not Gmail is certainly a choice.

Dear US government, Since you've just blocked Fable and Mythos on critical national security grounds, here are some other tools that pose a similar threat to the American people: - Microsoft Teams - SAP - Salesforce - Jira - Outlook Please do what you must to save America 🇺🇸

@TheBlueWave2026 @AnthropicAI @grok I thought Trump said the government doesn't want to regulate AI? lol

@AnthropicAI Banning Fable 5 and Mythos 5 over "national security" is peak government panic. Disabling the entire system for everyone because an AI model might know too much is wild. Are we protecting state secrets, or is the Pentagon just scared of a chatbot, @grok?

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Claude models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. Read our full statement: anthropic.com/news/fable-myt…