Transformer

An important update and apology from Transformer regarding our AI Campaign Finance Tracker: Today we discovered an error in our AI Campaign Finance Tracker. A bug in how our tracker deduplicates filings which has meant that we have undercounted spending from American Mission, part of the Leading the Future network, by about $9.6m, and double-counted about $1.4m of spending by Guardrails Alliance, a pro-safety PAC. The bug was introduced in late June, and most of the error built up in late October. Since the bug was introduced, our headline figures have therefore been undercounting “pro-innovation” spending and overcounted “pro-safety” spending. Post-midterms, our headline figures were $44.4m pro-safety vs $27.9m pro-innovation, or 1.6 to 1. The corrected figures are about $42.9m vs $37.0m, about 1.16 to 1. Pro-safety groups still outspent industry, but by much less than we reported. We know lots of people relied on our tracker as a source of truth in the midterms: as the editor ultimately responsible for it, I am very sorry that its numbers have not been accurate. The tracker has now been corrected and checked against FEC filings, so it is now accurate. Going forward, if we run a similar project in 2028 we will implement better review processes to make sure an error like this cannot happen again.

Our AI Power List with @NYMag doesn't just profile people — it also includes a bunch of mini-scoops. For instance: we're told that Jensen Huang, despite having invested a reported $5 billion in Ilya Sutskever's SSI, has not seen all the company's secrets.

News! OpenAI's written to Gov. Gavin Newsom recommending a bunch of amendments to strengthen SB 53. Includes a bunch of stuff on AI R&D and shutdown controls.

Anthropic's leaked AI bill is weak, fails to meet the moment, and makes the company's safety-first pitch ring hollow. The visceral negative reaction from across the AI safety community is, in some ways, evidence of a sense of betrayal: that the company that was supposed to be committed to their cause has, like all the others, let business interests win. transformernews.ai/p/anthrop…

I did not like Artificial, a jumbled mess that continually undercuts itself and lacks a coherent vision: transformernews.ai/p/artific…

Trump and Republicans want AI companies to regulate themselves. It’s a dereliction of duty that is failing Americans — and humanity as a whole. New from me in The Guardian: theguardian.com/technology/2…

As I said on @LBC: AI companies can no longer reliably control their systems.

This is a very clever evaluation design from UK AISI, basically seeing if AI models will misbehave the way they did in this summer's rogue AI incidents. When it comes to Astra, the answer seems to be "oh boy will they!"

Really great to be on @Channel4News tonight to discuss rogue AI agents...