Frankly, I give ZERO trust to the promises from OpenAI and Anthropic of not using my data for training. They never cared about copyright, using our session data for fine tunning it too juicy for them not to to dip their fingers on.
And thus, by restricting access to the latest models, the US government is giving a gift for the chinese labs, that will now receive lots of new users providing them with data related to real world software development.
Training is the least of my worries. They’ve got the single most valuable treasure trove of ad targeting data the world has ever seen. People type way too much personal stuff into these text boxes.
That trove that they “can’t” sell is going to get a lot of attention when they’re trying to hit quarterly numbers.
Heck, is it even against the current agreements for them to spin up their own Adsense network and serve personalized ads? They wouldn’t be selling your data, just Showing You More Relevant Ads™
The great thing is that many of the Chinese open-weight models can be run on US providers. Some promise not to retain data for training or other purposes. (Which I think is a credible promise most of the time.)
I think by limiting the usage other countries will start looking for alternatives, since you wont have access to the best models or you can be without any model in case the gov decides.
Is this article targeted just to an American audience? Because the US isolationism will only hurt them. Unless they truly have achieved AGI or ASI, the Chinese models soon will catch up. I'm pretty sure we will have an open weights frontier model this year.
AGI or no, if the US is right that cutting-edge models have national security implications, I doubt China will be any more enthusiastic about allowing people to release them open weights.
The real reason, afaik, that the US is trying to restrict access to SOTA models is that a very large component of USA tailored access injections and surveillance relies on exploits and weaknesses that these models will easily detect. Thus, it really is an export control issue, but it has nothing to do with offensive capabilities. Offensive capabilities always exist, but pervasive defensibility would upset the asymmetric advantage that attackers, especially the USA, currently have.
There are now Asian models coming , optimized and focused on cybersecurity defense at a high level. I suspect export walls will be a relatively moot point soon, because it is in chinas (and everyone else’s) interest to reduce US cyberwarfare dominance
LLMs are not great at creating exploits, but they are really good at detecting them. That asymmetry alone is enough to destroy the “offensive capabilities” narrative. Yes, mythos can find exploitable bugs, even write bench exploits. But real exploits require a good dose of human psychology, and most of the tools needed are off the shelf available anyway. You still need a real cybersecurity expert to effectively weaponize a zero day into a -deployable- exploit. If you don’t think so, just try to get Claude to help you with that. But it will gladly help you secure your systems.
What an LLM can do is inspect payloads, packages, and blobs at scale and find those exploits in a way that was wholly impractical before, so the asymmetric attack advantage is dissolved by strong LLMs.
The USA is trying to protect its cyberwarfare advantage, not protect against attackers. The exact opposite, actually. Porous security is a huge advantage to technologically advanced state actors.
And thus, by restricting access to the latest models, the US government is giving a gift for the chinese labs, that will now receive lots of new users providing them with data related to real world software development.
That trove that they “can’t” sell is going to get a lot of attention when they’re trying to hit quarterly numbers.
Heck, is it even against the current agreements for them to spin up their own Adsense network and serve personalized ads? They wouldn’t be selling your data, just Showing You More Relevant Ads™
There are now Asian models coming , optimized and focused on cybersecurity defense at a high level. I suspect export walls will be a relatively moot point soon, because it is in chinas (and everyone else’s) interest to reduce US cyberwarfare dominance
LLMs are not great at creating exploits, but they are really good at detecting them. That asymmetry alone is enough to destroy the “offensive capabilities” narrative. Yes, mythos can find exploitable bugs, even write bench exploits. But real exploits require a good dose of human psychology, and most of the tools needed are off the shelf available anyway. You still need a real cybersecurity expert to effectively weaponize a zero day into a -deployable- exploit. If you don’t think so, just try to get Claude to help you with that. But it will gladly help you secure your systems.
What an LLM can do is inspect payloads, packages, and blobs at scale and find those exploits in a way that was wholly impractical before, so the asymmetric attack advantage is dissolved by strong LLMs.
The USA is trying to protect its cyberwarfare advantage, not protect against attackers. The exact opposite, actually. Porous security is a huge advantage to technologically advanced state actors.
U.S. government will decide who gets to use GPT-5.6
https://news.ycombinator.com/item?id=48690101
U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations
https://news.ycombinator.com/item?id=48692995