Just a few days ago, the most heated debates about the merits of an open-weight class belonged to MMA enthusiasts. But after an open letter was released on Friday by more than 20 tech companies, open-weight AI models have become front and center in the discourse. For context: Open-weight AI models let users download and edit them, and run them locally. By contrast, the most popular US large language models are closed and proprietary, like Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 series. Why do we they care about this?Any connotation the public had with open-weight AI models was probably negative (at least before Denny’s intervened). Popular Chinese AI models, like those made by DeepSeek as well as newcomer Kimi K3 from Moonshot AI, are open-weight. The letter didn’t mention China…but it did enumerate the benefits of open-weight models despite bad press, including by extending access “to the AI economy.” AI will become scalable to use for billions of everyday tasks at “factories, hospitals, farms, classrooms, and main street businesses,” but only with the help of low-cost, open-weight models, the letter said. It also admitted there are risks with open models, like modifications being made to them beyond the creator’s control. Signer scrutinyNvidia and Microsoft were the heaviest hitters on the open letter. Another of the initial signers was Hugging Face, which was hacked by a proprietary OpenAI model after it went rogue…and it defended itself using an open-weight Chinese model after Anthropic’s Fable 5 couldn’t tell if Hugging Face was the hacker or not. The signers of the letter rose to 50 after Jensen Huang posted it on X. But those still don’t include Anthropic, Amazon, or xAI. (OpenAI signed after the letter gained traction.) Zoom out: The tech companies used the letter to urge the US government to avoid “premature restrictions on open models that stifle competition or drive innovation overseas,” following interventions it’s made in domestic AI companies.—HVL | 
.png)
