6 months to live for open models
Interconnects Nathan Lambert
White House talks could ban powerful open-weight AI models within six months. Open models have no big lobbyist fighting for them — that's the real risk here.
Based on reporting by Interconnects, Nathan Lambert — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Something unusual is happening in Washington right now, and it could reshape open-source AI faster than any previous scare. Multiple sources are pointing to White House discussions around a new executive order meant to manage open models. Nothing is official yet, and the early scope reportedly covers Chinese-origin models and government use only — but that's exactly how these things tend to start small before they don't.
The clearest signal of how fragile the open ecosystem's position is came out of a June 9th meeting, where a Reflection AI representative pushed for open-source models to get exemptions from a proposed licensing framework based on their actual capabilities. It's a reasonable ask on paper. But it also reveals the problem: right now it's Chinese labs like DeepSeek holding the lead among open models, while Reflection itself hasn't even shipped a public model. There's no single, well-resourced champion standing up for open weights the way closed labs stand up for themselves.
The likely outcome, given the pace of progress, is a ban or indefinite delay on any open-weights model that meaningfully clears the capability range of GPT 5.5, Claude Opus 4.8, or GLM-5.2 — and that could land within six months. What makes this messier is that two separate policy fights are getting fused into one narrative: distillation concerns and frontier-capability concerns. They're different problems needing different answers, but together they're fueling a single, gathering push toward restriction.
Much of the distillation alarm traces back to Anthropic, which has publicly detailed foreign companies paying for and using its API, then cut off that access and followed up with policy recommendations light on technical evidence. The irony is hard to miss: Anthropic argues its models, like Claude's Mythos, are powerful enough that open equivalents should be restricted, yet Mythos itself was accessed without authorization by outside users during its most locked-down private beta. If APIs were the obviously safer alternative to open weights, that shouldn't happen.
The most realistic pressure valve here isn't a policy fix — it's a product one. If a U.S. company such as Microsoft or Meta released an open model at a comparable capability level, it would undercut the framing that only Chinese labs are pushing the open frontier forward. Absent that, the loose coalition of companies and developers who actually build on open weights needs to start organizing now, because right now nobody with real leverage is making their case.
My take — AI-written commentary, not fact-checked reporting
This has less to do with safety than with who gets to gatekeep access to powerful models, and Anthropic's pattern of dropping warnings without evidence looks a lot like textbook regulatory capture dressed up as concern. If a company's own flagship model can be accessed by outsiders during a locked-down private beta, that company doesn't get to claim APIs are the safe alternative to open weights. Banning a handful of Chinese open models won't stop distillation or bad actors — it'll just hand the entire open ecosystem's momentum to whichever country doesn't flinch.
Read more about this at: Interconnects
