Box CEO Aaron Levie argues restricting frontier AI releases will fail to protect US dominance as open models proliferate

@levie @paulg Betting against people is generally a bad bet. Esp when 50% of our AI researchers are Chinese.

@levie China is doing with AI models what the United States did with the Internet: open, affordable, and neutral.

@levie MSFT and the feds better be pushing out 'security updates' before the French and Germans find out how many security loopholes were left in for the American spooks

@levie The real opportunity is to have a "Frontier Certification". Have available a Mythos/GPT certification model scanning and repairing your code to the highest standards. Either a SaS or a free service funded by the US government by an AI tax. Certification Star on your website.

@levie 🎯

@levie You act like the US permanently banned these models rather than slowing down the rollout by a few weeks. I feel like there have already been many freakouts and delays and BS around new models for years and there will be many more. Nothing's perfect.

@levie @paulg It’s at the level of flopping in the NBA.

@levie Indeed

@levie @paulg People living outside US and non US are fed up with US Admin and tech erratic selfish behavior, from today I flipped all my agents running on Hermes to Zhipu glm 5.2 https://z.ai/blog/glm-5.2 Fuck closed Ai llm 🖕, ❤️ 🇨🇳 open source LLM models

@levie Heavy +1, this is going to be a nightmare for security teams once Chinese models improve to the levels of mythos. We are not too far - may be a month or 2.

@levie @paulg Just like everything else, what China can’t get cheap enough, they will steal.

@levie Feels like we're speed-running the "no one can ever build a car like Detroit" playbook again, but this time instead of it taking a couple of decades it's going to take a year or two.

@levie Gatekeeping only works if your competitors agree to play by your rules. China won’t.

@levie @paulg When cyber AI hits open availability, security tooling economics change fast. Expect a wave of composable agents for continuous verification rather than periodic scans, worth exploring how that integrates with cloud-native infra at scale.

@levie China’s not waiting, and neither should we.

@levie I wish more people in this industry, especially in the Labs, had this level of clarity in their thinking. Sadly, my gut tells me it has nothing to do with brains and a lot to do is a distraction from the real goal which is stiffle competition via oss bans.

@levie This is exactly what meant when I said AI will create new jobs. I’m not in the camp that robots will be running the world leaving humans unemployed.

@levie @paulg Look at what Regulation did to AI in Europe. Why would anyone want to follow this path?