“While we were able to freely share our highest-level conclusions about risk, Anthropic redacted a majority of the evidence we found most informative for these conclusions.”
The headline is that Anthropic now builds capability about twice as fast as it did a year ago, mostly by using its own AI. The buried detail is that you can’t check any of it. The company being audited redacted most of the evidence, and the auditor says so in the first paragraph. That’s not independent oversight. It’s a press release with a third party’s logo on it.