issue 01: access denied

the week Washington reached into the model layer. one export-control letter, and the most powerful model on earth went dark for everyone.

what happened

• The US Commerce Department ordered Anthropic to suspend Claude Fable 5 and Mythos 5 on June 12, three days after their June 9 launch. The directive bars access by any foreign national, inside or outside the US — so Anthropic disabled both models for every customer to comply. (Anthropic, Axios)

the standoff

• The White House says Anthropic refused to fix it first. David Sacks, co-chair of the President’s Council of Advisors on Science and Technology, says the administration warned Anthropic that the model had been jailbroken and asked CEO Dario Amodei to patch the bypass or de-deploy it. He declined; the export control followed “reluctantly.” (Tom’s Hardware, CNBC)

the trigger

• A reported jailbreak, and a foreign breach. The order followed a demonstration that the model’s safeguards could be bypassed, with a foreign group reportedly having accessed it — enough to alarm the administration. Anthropic reviewed the report and says the capability shown is “widely available from other models,” including OpenAI’s GPT-5.5. (Fortune, Anthropic)

why it matters here

• Foreign nationals — Canadians included — are now first in line to be cut off. Model access just became an export-control surface, revocable in 72 hours. Canada’s ~$2B “AI for All” strategy, with its technological-sovereignty pillar, stops reading like national-pride PR and starts reading like a procurement requirement. (CBC)

worth a longer read

write the loop that stops. Our launch field guide argues for an LLM judge you can swap on a config line rather than one you hard-wire — written days before a government order proved exactly why you don’t bet a system on a single model. (read it)

§01 · 60°N · the weekly canadian AI digest · northofthehype.ai