Ask HN: Model access depends on citizenship. What should Non-US founders do?

LLMs are at this point a necessity to develop large scale systems - we couldn't do what we do without them. Our team is worried by the developments of Mythos/Fable getting restricted and the news about GPT 5.6 access being gated by the government customer-by-customer.

The part that complicates the obvious answers: the Anthropic directive keyed on the nationality of the person accessing the model, not where the company is incorporated; even their own foreign-national employees lost access. So "just flip to a Delaware C-corp" doesn't clearly solve it for a team like ours.

This is all new and we don't think anyone has real answers yet. We want to hear what others think might work. A few directions we've floated, none of which we're sure about:

- Hoping that at least one frontier model provider remains accessible - An open-weight/non-US floor (GLM/DeepSeek/Qwen) - Hoping for frontier access via Bedrock/Vertex/Azure rather than first-party APIs

Our concern is that the regional capability difference gets worse over time. We'd all be happy to naturalize to US citizens, but is this even possible - and can we do so before the US compute lead combined with regional restrictions locks us out of being competitive in software production?

What are we missing? How are other non-US national founded startups mitigating this (emerging) risk? Are we overestimating it or where could we be wrong?

[Context: South Africa, Pre-seed, SMB AI Software Infrastructure, 100% South African Team]

▲

verdverm 13 hours ago | parent [-]

Caveat, I'm US based, but still hedging by buying hardware and open weight (per-token) vendor. I can no longer trust Big AI nor my government (both parties are trouble). Letting the rich and powerful decide who can access what models is not a future I want. Open weight is the way forward for most people. Frontier will be for niche applications like bio engineering and other niche that are less language oriented and require specialized models.

This was on HN a day or so ago

www.anildash.com/2026/06/23/fight-ai-platform-war/

▲

recsv-heredoc 13 hours ago | parent [-]

Great article. This is exactly what we're doing from a product perspective.

What if the frontier-minus-6-months assumption does not hold? The US has 5x the AI Capex of China, and 10x the EU. Assuming AI is compute limited (we certainly seem to be given the RAM crisis) - wouldn't it be reasonable to assume frontier models are likely to continue to pull ahead?

	▲	verdverm 11 hours ago \| parent [-]
		I'm not sure capex is the best metric, PPP make's China's money go further. Their techsci funding (subsidizing) produces more per unit. Also consider that Ai needs electricity. China has more than 2x the generation capacity and is building faster than anyone else. Here in the US, we are facing pushback that will slow things down. Then there are the model capability questions. (1) Frontier models are being restricted (2) eventually models will be good enough for most tasks. This later point is why I believe the most of us will eventually have unlimited plans like our mobile data. Switching model/providers is very low effort as well. We already have tools like LiteLLM and GoModel that provide a single endpoint to this capability.