ARCHIVE
Every article published in November 2026, across all sections (8 articles).
The evaluator METR estimates that Anthropic's models are likely improving at about twice their earlier pace. OpenAI introduced a benchmark for reading lab data, and Censys reported exposed consoles on 1,465 hosts for a tool linked to attacks.
OpenAI released a new cybersecurity model to a select group of approved customers and published pledges to the communities that host its data centers. Anthropic said it is starting a US beta that lets Claude read bank and card accounts without moving money.
OpenAI announced it will buy a maker of tools for checking code, and Google published the method it uses to evaluate AI for health.
Anthropic's amended IPO filing put numbers on its revenue and its spending commitments, while Google and Apple announced changes to how children and teens reach AI.
A day about who carries the cost of running AI, in three places at once: news articles, electricity and usage allowances.
A day when people outside the tech industry draw the lines around AI. A federal court sent a lawsuit over a data center's power plant forward, a mathematician asked his field to rethink how it gives credit, and California voters decide on an environmental-review measure on November 3.
Last week in AI (Oct 26-Nov 1), Anthropic's IPO filing and Big Tech's earnings put the money behind AI on the record, and outside researchers took on the job of checking proofs that AI models had produced.
A day about how AI reaches people. A tampered extension arrived through an official distribution channel, and Apple will bring its AI features to China with a local model built in.