Samuel Roland Profile picture
Infrastructure @joinFAI, Law @GMU, AI dabbler on the side. Per conatus et fidem ad astra simul.
Mar 24 18 tweets 5 min read
This is going to be a live tweet thread on the Anthropic PI hearing starting at 4:30.

For those who have not had the chance to catch up, Judge Lin's order with the discussion questions is below.

1/x

storage.courtlistener.com/recap/gov.usco… Judge Lin starting with tentative thoughts, says Pentagon clearly has the ability to find a more permissive vendor, but the case is not about that.

Case is, in her view, about the failure of tailoring here. Quotes "attempted corporate murder" @deanwball
Feb 28 6 tweets 2 min read
Since I'm up late working on this anyway, what the hell, time for a tweet thread on why @sama's announcement might help save Anthropic from the worst outcome here. 1/x There are two statutes here that could support a "supply chain risk" designation. Both require a finding that "less intrusive measures" aren't reasonably available (41 U.S.C. § 4713(b)(3)(B); U.S.C. § 3252(b)(2)(B)).

According to OpenAI, the government just showed that it can accept the same restrictions through negotiation. Under any definition of the term, that is a less intrusive measure than designation. 2/x