Georeactor Blog

I have no inside insight into this but some thoughts -  https://12gramsofcarbon.com/p/tech-things-there-is-a-massive-shadow - is an interesting piece on the absurdity around the government's Fable ban, and other weird developments such as Bernie Sanders being "the most AI pilled member of Congress".
Note: I think Sanders is realizing the big AI IPO moment is now, and trying to connect to the real discussions which leftists were having about automation and unemployment during the Andrew Yang / UBI era. Plus Sanders has been a longtime advocate of employee ownership of companies (in 2019 the Cato Institute called it a "Stock Confiscation Plan"), so it's not surprising that he thinks that the US should take a stake in the AI IPOs, especially if the administration is already considering it. It's just uncomfortable for people who expect a unified leftist front against AI and one of their heroes is on a whole other planet.

Similarly there is no firm direction on the right when it comes to AI policy, just warring factions, including a pro-business / deregulation faction which often gets what they want. Despite statements from the first term, this administration agreed to deregulate cryptocurrency, including disarming the SEC and pardoning people and a corporation.
The 'culture war' faction criticized Biden's executive orders as attempts to make AI 'woke', and celebrated them being revoked.

But a year later, AI's abilities are making some people nervous. Is there any line where the government would intervene? This recently played out with the lead-up to an executive order which was halted in May after leaks (most likely from the deregulation and anti-woke factions) that Trump's order would end up similar to Biden's. In the end they were able to get an industry-friendly voluntary order on June 2nd.

There's a theory that maybe the Fable ban is a continuation of Anthropic's dispute with Trump and the Pentagon back in February. Totally disagreeing with cause->effect from the Substack post, I believe that members of the 'anti-woke' faction -  alarmed at Anthropic's talking points around Effective Altruism, Claude Constitution, and their cultural interview - were what initiated the dispute. I know that I couldn't explain those alongside what they're supposed to deliver to the Pentagon and NSA.

It's possible that Anthropic's hype in the past two months of Mythos has been enough to scare someone influential within the administration that this tech shouldn't be public. And also possible that after February, Anthropic no longer has a direct line to Trump to wave off these concerns (Sam Altman told the New Yorker "You can just, like, call him.").

Anthropic's response so far is appealing to the deregulation faction ("If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers"). They also point out that the government showed them minor vulnerabilities, implying that they need allies familiar with present LLM abilities to step up.
I could only read this as an intimidation campaign against Anthropic if it lasts long enough to disrupt their IPO. It seems more likely to me that the same people who wrote the H-1B orders and other sudden, unimplemented policies were told "block other countries from using that AI" and didn't expect this to be so disruptive.