Fund U.S. open models

Federal push for U.S. open-weight models. The government funds and promotes open-weight American models so the world builds on U.S. technology.

Grows lead by~9 daysof America's lead over China, over 3 years

Raises p(doom) by~0.15%from a 5% starting estimate

Holds up○ Training○ Lab's own use○ Public release

Enact it on the menu See the findings

What it does

The government would fund, compute-subsidize or otherwise favor American open-weight models so U.S. models, not Chinese ones like DeepSeek or Qwen, become the world's default. Anyone, including rivals, could download and modify the strongest U.S. open models.

Applies to: Federal research funding and procurement.

The government funds and promotes open-weight American models so the world builds on U.S. technology.

Where things stand

U.S. open models exist (Ai2 OLMo, OpenAI gpt-oss, others) but Chinese labs lead open-weight releases. Federal support is small grants and DOE science models, not a frontier-scale open lab.

Why it gains ~9 days

Open models speed U.S. research a little. They also hand the same models to China, which matters if China depends on U.S. models.

Biggest unknown: Whether U.S. open frontier weights help America more (global adoption, research) than they help China and bad actors (free capabilities, no recall).

Why it raises p(doom) by ~0.15%

Not a safety law. Funds and promotes American AI models that anyone can download.

Once weights are public, safeguards can be stripped out and anyone can use them. Open models also help safety research, so this is debated.

The strongest case that it costs more

Supporters argue open American models set global standards and keep the world from building on Chinese open models.

The debate

For

Nathan Lambert, Ai2 (ATOM Project) 2025: argues America needs at least one open-model lab with 10,000+ H100s to match China's open labs.
White House AI Action Plan July 2025: 'We need to ensure America has leading open models founded on American values.'
Nvidia (Jensen Huang) and industry letter 2026: Nvidia, Microsoft, Google, Meta and startups urged the administration not to restrict open models.
U.S. Department of Energy Aug 2026: Genesis Open Models, releasing weights 'to galvanize the scientific and AI communities'.

Against

Anthropic (Dario Amodei) 2026: warns open weights cannot be recalled or patched once released and that China could use models for military advantage.
NSA, CISA and FBI Sept 2026 advisory: six Chinese AI firms ran large distillation campaigns on U.S. models, showing how capabilities leak to rivals.

Sources

Interconnects: The White House's plan for open models (July 2025): Action Plan open-weight language and analysis.
DOE: Genesis Open Models Initiative (Aug 7, 2026): First government-backed open-weight model program; Genesis-Science-1 with Arcee.
FedScoop: NSF, NVIDIA partner on open AI models for science: $75M NSF + $77M NVIDIA to Ai2.
Axios: Trump AI framework excludes open models (Aug 4, 2026): Open-weight models outside voluntary pre-release review.

Rough starting points, not precise forecasts. Lead costs assume China doesn't depend on U.S. models, the case least favorable to safety laws, and count 3 years. On the menu you can change every assumption and put in your own numbers. Last priced 2026-09-26.