Anthropic reportedly withheld Mythos 5.1 from UK AISI pre-release testing, testing cross-border AI safety cooperation.
Financial Times : Sources: Anthropic declined to submit Mythos 5.1 to the UK AISI for pre-release testing, prompting UK fears that US AI labs are aligning with US protectionism — Decision to exclude Aisi has prompted fears inside British government of protectionist shift among tech groups.
NSA, CISA, and FBI warn that Chinese AI firms including DeepSeek are conducting industrial-scale model distillation.
Reuters : The NSA, CISA, and FBI issue a joint advisory warning that Chinese AI companies, including DeepSeek, are conducting “industrial-scale” distillation campaigns — The U.S. government on Tuesday accused Chinese artificial intelligence companies of maliciously copying technology …
Suno v6 was trained with WMG and BMG and introduces royalties for labels and publishers when models are used.
Bloomberg : Suno rolls out v6, a new line of AI models trained in partnership with WMG and BMG, and says it will pay labels and publishers royalties when models are used — - New model was trained on music from Warner Music Group — Users won't be able to reference specific artists in prompts
OpenAI details how GPT-5.6 Sol is being used to operate and analyze quantum-computing experiments.
Dwarkesh analysis attributes 2019–2025 pretraining compute-efficiency gains chiefly to better data, not model design.
Dwarkesh Podcast : Analysis: from 2019 to 2025, gains in pretraining compute efficiency came mostly from data improvements rather than model improvements — Breaking down 6 years of pretraining progress into data vs model improvements — How much of the rapid progress in AI that we've seen over the last few years 1 …
Anthropic researcher Jacob Coxon is leaving AI over fears labs are racing toward systems they cannot control.
Amrith Ramkumar / Wall Street Journal : Anthropic researcher Jacob Coxon says he is quitting the AI industry over fears that tech companies are racing to build systems they won't be able to control — Concerns are rising inside AI labs that competition is pushing tech companies to race toward self-improving models that risk spiraling out of human control
New OpenReview research finds LLMs develop novel social biases through adaptive exploration.
Google plans a €13B-plus, two-year Finnish data-center buildout—its largest European investment.
Kirsi Heikel / Bloomberg : Google plans to invest €13B+ in Finland over two years, its biggest investment in Europe to date, to build three new data centers and expand an existing one — Alphabet Inc.'s Google is planning its biggest investment in Europe, an artificial intelligence infrastructure build out worth …
More than 10 US states are rolling back data-center tax breaks amid escalating backlash to AI infrastructure.
Wall Street Journal : Over 10 US states have rolled back tech giants' tax breaks, which top $1B/year in some states, as many others propose similar bills amid a data center backlash — Amazon, Meta and Google risk losing decadeslong exemptions due to a backlash against their facilities
Microsoft’s September Patch Tuesday fixes roughly 972 vulnerabilities, a record monthly total.
Dan Goodin / Ars Technica : Microsoft's September 2026 Patch Tuesday fixes a record ~972 vulnerabilities, bringing its total flaws patched in 2026 to 2,760, more than double from 2025 — Microsoft's patch for September is a doozy, with a record number of roughly 972 vulnerabilities fixed and 112 of them meeting the high critical-severity threshold.
Gimlet Labs raised $300M after citing possible $100M-plus annual OpenAI spend, which OpenAI says has not begun.
The Information : Source: Gimlet Labs raised its $300M round after telling investors OpenAI may spend $100M+ annually on its services; OpenAI says it's not a paying customer yet — Gimlet Labs said late last week it had raised $300 million at a valuation of about $3 billion, in a round led by new investor Andreessen Horowitz.
Inception Labs launches Mercury 2.5, a new model release from the diffusion-language-model developer.
A GitHub project runs the 2.8T-parameter Kimi K3 at 1 token/s on a MacBook Pro by streaming weights from four SSDs.
Benchmarking finds 4-bit Qwen3.8 27B quantization remains robust while 1-bit quantization collapses.
OpenUI introduces OUI-1, positioning it as the first generative-UI model.
This interactive tool visualizes LLM attention, making token-level attention patterns inspectable.
MIT’s 2025 paper develops a microeconomic framework for artificial intelligence.
C* proposes unifying programming and formal verification within C.
Copperhead is a Cursor-like AI-assisted environment specifically for circuit-board design.
Ars Technica reports an LG TV scanning its local network for third-party phones and other devices.
Harvard researchers report a model that predicts most suicide attempts one week in advance.
UK police raise concerns that Meta smart glasses could enable covert recording of officers.
China’s securities regulator is reportedly tightening IPO approvals for humanoid-robot startups after Unitree volatility.
The Information : Sources: China Securities Regulatory Commission is informally tightening IPO approvals for humanoid startups after a volatile debut by industry leader Unitree — Chinese regulators are tightening the approval for humanoid startups seeking to go public after a volatile debut …
Apple’s first foldable is reportedly named iPhone Duo, priced from $2,000, and may ship as early as October.
Mark Gurman / Bloomberg : Sources: Apple's first foldable will be called iPhone Duo, with prices starting at $2,000 for 256GB of storage and sales starting as early as October — Apple Inc. is set to debut a roughly $2,000 foldable iPhone Duo on Wednesday as part of a wide-ranging product event, which kicks off at 10 a.m. local time from Cupertino, California.
FreeBSD 14.5-Release is now available from the project’s official site.
A 27.5KB, language-agnostic WebGPU syntax highlighter demonstrates compact GPU-based text rendering.
An interpreter became 17% faster by replacing a Rust enum with a packed 64-bit word.
A Reddit discussion examines claims that research was stolen from private Codex chats.
This essay argues that AI has a discovery problem, focusing on how users find useful models and tools.
Y Combinator launches its Early Access Network for connecting startups with early users.
Sarah Wynn / The Block : Malone Lam, a 22-year-old Singaporean national, pleads guilty to leading an international social engineering scheme to steal and launder $245M in cryptocurrency — - Prosecutors say Malone Lam, who went by aliases “Anne Hathaway,” “$$$” and “King Greavy,” led the operation.
Evan Hubinger / @evanhub : Anthropic's Alignment Science lead says there is a “>10%” chance AI could kill all humans within the next decade and worries about recursive self-improvement — Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.
Hayden Field / The Verge : In an expanded class action lawsuit, a group of Claude subscribers says Anthropic deceptively advertised the limits of its Max subscription tier — Anthropic says power users are key to its business — it's prioritized them even when it means cutting off other popular applications, like OpenClaw.
Alex Heath / Sources : Q&A with Mark Zuckerberg on Muse, the data center backlash, Meta's recent youth-safety settlement, privacy concerns about Meta's smart glasses, and more — Meta's CEO joins me for an in-depth conversation on episode two of the Sources podcast. — When Mark Zuckerberg laid out his …