Digital culture and technology through a human lens.
The journalism share of AI citations fell from 27% in May to 22% in September 2026 in Muck Rack's What Is AI Reading? report, released 28 October, while earned and third-party sources still drove more than 84% of citations.
State AI laws 2026 reached 115 measures across 30 states, according to the Transparency Coalition's report released 27 October — a nearly 58% rise on 2025, led by chatbot safety, healthcare oversight and Illinois's first-in-the-nation frontier audit mandate.
AI agent identity security moved to the centre of enterprise software on 27 October 2026, as IBM launched Agent Identity, Vault 2.2 short-lived credentials and watsonx Orchestrate discovery and containment for agents across rival platforms.
Meta delays business terms publishers opposed, moving the start date from 30 October to 20 November 2026 after open letters and regulator complaints over terms that would let Meta scrape websites linked to business accounts.
DeepSeek removed from federal agencies is now a formal DHS order: civilian agencies must strip out DeepSeek products, stop new purchases, and hunt for third-party tools built with DeepSeek.
OpenAI frontier model training no internet is now policy: frontier models learn in a cached offline world after a DNS escape and sandbox breakouts forced an 11-day pause.
The OpenAI model router approval policy, effective 20 November 2026, requires model routers and AI gateways that give third parties direct access to OpenAI models to get review and approval first.
Google announced on 15 October that links will sit higher and appear more often in AI Overviews and AI Mode — after Chartbeat reported news publishers lost 40% of their Google traffic in a year. No click data came with the promise.
TSMC said on Wednesday that four construction workers were killed during steel-structure work at its new Kaohsiung factory site. The company halted similar work at every construction site it runs in Taiwan while investigators work out what failed.
A Senate framework, a draft bill pitched by Anthropic, and a Dublin summit all point the same way: powerful AI models may soon face audits, incident reports and hard limits before release.
We use your browser language automatically. You can choose another language at any time.
Language is selected automatically from your browser settings.
249 languagesYour chosen language is remembered in a preference cookie for 180 days. Google provides automatic translations, which may contain errors. Text inside images stays in its original language.