MeusterMint

If you’re an AI investor, the next “alpha” is literally on our face right now…. So today an AI hacking group just went after 400+ banks with almost no humans in the loop. A team of agents scouted targets, stole passwords and broke into accounts. When one model refused, the system simply switched to another and the security firm that caught it says offense is now well ahead of defense…think about that for a second. If you’re in the Trump admin or any government official in the western world or allied with the United States, you’ve got a mandate to protect your “sovereignty” and citizens. This is obviously a major national security threat and I’ve tried to talk about it as much as I can on national TV and on X. When this attack was reported, Anthropic and the other US labs shut the hackers out. Their models run on their own servers, so they can see abuse, trace it and pull the plug. That's the whole ballgame imo which takes me to my next point… You can't do any of that with an open model. Once the weights are downloaded, they're out forever. There's no off switch, no audit trail and no one to call. A model sitting on someone's hard drive can't be revoked…which nobody wants if such attack happens. And btw this is only the beginning…think about where this goes from here… Today it's passwords at banks in Latin America. As these models get smarter, the same playbook works on hospitals, power grids and payment systems. No government can tell its citizens that frontier level intelligence will stay freely downloadable by anyone (even by adversaries or terrorists), anywhere, with zero oversight. Especially when some of the most capable open models come from labs outside US jurisdiction. So I think we're heading toward a "know your customer" era for AI, the same way banks have to know who they're moving money for. Expect licensing for the most powerful models, real accountability for whoever serves them, and a ban on open weight Chinese models above a certain capability level of all as a whole. It’s not so crazy to think that’s going to happen because early this summer the Trump admin banned an “American” made fable mythos model from Anthropic for its godly hacking capabilities. So if they’re going to do that for American models what makes you think they won’t do it to Chinese models? It’s a no brainer… Every one of those rules points the same way…toward labs that can control, monitor and shut down how their models are used. That's the new moat and it's one money can't easily buy. Compliance, monitoring and abuse detection take years to build. Enterprises and banks will start asking a simple question before they plug in any AI: who's accountable if it goes wrong? Anthropic, OpenAI, SpaceXSI and Google can answer that. A model downloaded off the internet can't… And as public markets investors there's an infrastructure angle. If frontier AI has to run where it can be watched, it runs in datacenters, not on random machines. Centralized inference means more demand for the hyperscalers, the chips $AMD/ $NVDA and the memory $MU and $SNDK behind them and as of today’s closing prices we think they’re totally undervalued and misunderstood. And to be fair, open models have real value and no regulation will be simple. The hackers used US models too…the difference is that the US labs could cut them off. That's exactly the point I’m trying to make. Safety after today will starts to look like the most valuable feature a model can have…limit up!