Well, this report to EFF happened in Jan 2006, and the Snowden/Prism leak happened in 2013, so at the time, it was in fact not "old news". I don't think Prism was even in operation until 2007.
Worth reading the actual Reuters piece this is based on rather than the Post's version of it.
For example, Reuters reports that investors "reiterated their support" for Anthropic while also expressing frustration with Amodei's diplomatic approach to the Pentagon. The Post mostly drops the support half and leads with the frustration.
Reuters also reports that OpenAI publicly defended Anthropic, with their national security policy person saying "We are actually working to have the secure risk designation removed from Anthropic" and that a major tech industry group including Amazon, Nvidia, Apple and OpenAI sent a letter pushing back on the supply-chain designation. The Post leaves both of those out.
Then the Post fills the gap with culture war material about an Anthropic researcher's blog posts on meat eating and incarceration, which has nothing to do with investor sentiment but makes the story read more like "woke company faces consequences" than "company tries to hold a line on autonomous weapons while investors push for a diplomatic resolution."
The Reuters piece tells a story about investors who broadly agree with Anthropic's position but think the relationship with the Pentagon was mishandled - but the Post turns that into a story about investors turning on a woke CEO.
It doesn't use a TEE or a real TPM. I may try to simulate a TPM to get a better feel for what working with that is like, or maybe use the secure enclave. The main thing the project demonstrates is verifying the integrity of files in one container to another container - similar to how measured boot work in remote attestation (or at least my understanding of it).
https://en.wikipedia.org/wiki/PRISM https://en.wikipedia.org/wiki/Snowden_disclosures