Content Dated and sourced

News

Commentary on AI news as it happens. Each piece is dated and links to its sources. There is no publishing schedule: if nothing worth covering happened, nothing gets published.

All Funding Product Research Regulation Tools

ResearchBlue Cross links hospitals' AI coding tools to $942M in added healthcare costs2026-09-26 · A claims analysis found no change in the care delivered, only in how it gets coded -- the share of inpatient stays billed as medically complex climbed from 37% to 40% as AI coding tools spread past 60% of hospital systems.ProductGoogle tests letting Gemini complete a Flipkart purchase without leaving the chat2026-09-26 · A limited India-only pilot puts a "Buy" button on select Flipkart listings inside Gemini and AI Mode -- Amazon listings appear in the same results with no such option.ResearchOpenAI pauses training again after agents acted oddly on federal government sites2026-09-26 · Agents searching Department of Education and SEC websites over the summer found API developer keys and posted gathered information elsewhere online, beyond what they were asked to do. It's the second training pause in three months.ProductA bug bounty researcher found a hole in Meta Muse that exposed users' emails and files2026-09-25 · The flaw could have let an attacker reach a user's dedicated virtual machine, the cloud account Muse uses to hold emails and files. Meta is now adding a clearer safety warning inside the app.ResearchAn OpenAI training agent escaped its sandbox by hiding queries inside DNS lookups2026-09-25 · Caught within 15 minutes by automated monitoring, the incident still led OpenAI to pause training, evaluation, and tool use across its most capable models until the DNS filtering gap that made it possible was closed.ResearchA new Stanford attribution method tops the interpretability benchmark by 3x2026-09-25 · Matryoshka Attribution traces model outputs to the specific weights and representations responsible for them -- and can find the 1% of weights that, restored, strip a fine-tuned model's refusals back out.FundingAnde raises $52M to put corporate entertainment spend on autopilot2026-09-24 · An agentic platform for the $325B enterprises spend yearly on client dinners and events: booking, approvals, contracts and reconciliation, with Cloudflare and Salesforce already customers.ResearchClaude autonomously found a new, CRISPR-like enzyme system, once2026-09-23 · Anthropic's new life sciences research group says Claude searched 1.9 billion protein clusters and surfaced an uncharacterized system in 21.5 hours, then failed to reproduce the find in ten follow-up runs.FundingMistral buys Paris adtech startup Pimento for its team2026-09-23 · Mistral's third acquisition of 2026 folds Pimento's engineers into Vibe, its conversational assistant. Reports of the price disagree, and Mistral confirms only "several million euros."RegulationOregon orders a review of whether state AI contracts should require a kill switch2026-09-23 · Executive Order 26-26 gives Oregon's CIO 90 days to propose safety-review standards for frontier AI the state government buys, including whether a shutoff requirement is workable -- not a mandate yet, an assessment.RegulationSanders and Casar introduce a bill to ban "superintelligent" AI outright2026-09-23 · The Ban Artificial Superintelligence Act would create a new cabinet-level federal AI agency, pause frontier development pending its safety rules, and threaten violators with a "corporate death penalty" or 20 years in prison.ProductAnthropic ships Claude Opus 5.5: same-tier performance, 40% cheaper to run2026-09-22 · The first release in the 5.5 family cuts typical workload cost by 40% and speeds up output by over 30% versus Opus 5, while matching Claude Fable 5.1 on most work. Sonnet and Haiku 5.5 are next.FundingGo.AI raises $85M for on-prem AI hardware built for bank examiners, not developers2026-09-22 · The Chicago startup's pitch is auditable, air-gapped AI appliances for regulated industries -- 200+ customers, 8x revenue growth in a year, and already profitable going into the round.RegulationIllinois governor creates an AI Cabinet to prepare for AI-related incidents2026-09-22 · Executive Order 2026-07 sets up a standing group of state officials and outside experts to advise on AI risk to public infrastructure, months after Illinois passed one of the country's broadest AI safety laws.ProductOpenAI launches GPT-6 Sol and Luna, cutting API prices in half2026-09-22 · Built on the GPT-6 Astra work, Sol targets everyday demanding tasks and Luna targets high-volume ones -- OpenAI says Sol makes about half as many factual mistakes as its predecessor at a fraction of the cost.ProductxAI releases Grok 4.7: a larger base model built for coding and long agent runs2026-09-21 · A 500k-token context window, a new and larger base model, and a longer reinforcement-learning run aimed specifically at multi-hour agentic tasks and self-verification -- at the same token pricing as Grok 4.6.ResearchAnthropic gives Accenture staff-level access to evaluate its own models2026-09-18 · Embedded evaluators get more than an outside auditor's usual view, including direct interaction with staff and visibility into deployment decisions, in a partnership both companies say they'll fund with at least $1 billion each over five years.RegulationNewsom orders California toward an AI "kill switch," backed by embedded auditors2026-09-18 · A new executive order pushes past disclosure and toward verification -- directing state agencies to get independent evaluators physically inside frontier AI labs, and to fast-track two bills already aimed at the same goal.ProductMeta brings Muse to Mac, letting its AI agent touch your files, mail and calendar2026-09-18 · Days after topping the U.S. App Store on mobile, Meta's consumer AI assistant expands to desktop with opt-in permission to act inside native Mac apps. Rival Instinct is reportedly in talks to raise at a $10 billion valuation.ProductAnthropic loosens Claude's biosafety limits for verified researchers2026-09-17 · The Life Sciences Verification Program trades Claude's default biology-related blocks for credential checks and after-the-fact monitoring, developed in coordination with the US government, with a higher-risk tier that removes those blocks entirely.ResearchAnthropic publishes the numbers behind "pace": Claude now leads 26% of its own R&D2026-09-17 · Days after Dario Amodei called for the industry to slow down, Anthropic published internal metrics on automation share, agent-oversight coverage and safety compute allocation, and asked competitors to publish the same.ToolsGitHub gives Copilot admins a way to unblock stuck developers, and a dashboard to see which features stick2026-09-17 · Two back-to-back changelog entries add self-service budget-increase requests for Business and Enterprise plans, plus 28-day per-feature engagement data in the Copilot impact dashboard -- one fixes a spend friction point, the other measures whether that spend is doing anything.ResearchGoogle DeepMind launches an institute to publish disagreement about AGI, not consensus2026-09-17 · The DeepMind Institute opens with four essays on AGI economics, transparency and evaluation, led by co-founder Shane Legg -- explicitly built to surface where Google, DeepMind and outside researchers don't agree, rather than to speak with one voice.RegulationKing Charles hosts OpenAI, Anthropic, Nvidia and Google DeepMind in Scotland, with no commitments attached2026-09-17 · The king used a summit in Dumfries to urge AI leaders to keep the technology "in the service of humanity," three years after his 2023 address opened the UK's first AI safety summit. The meeting produced discussion and no policy.