Anthropic's alignment lead says the company has no plan for superintelligence
Evan Hubinger, who runs alignment science at Anthropic, publicly said the company does not yet have a plan to solve alignment for superintelligence and is not clearly on track to get one, and put the chance AI kills all humans within a decade above 10%.
Anthropic is building predictive surveillance that tracks activists, an investigation reports
The American Prospect reports Anthropic is assembling a threat-intelligence operation that monitors protest activity and attempts to predict incidents before they happen, and that a company job posting lists "activism" alongside terrorism and nation-state threats.
US agencies name six Chinese AI firms and say copying American models is their core strategy
A joint NSA, CISA and FBI advisory says China-based AI companies including DeepSeek, Moonshot, Alibaba, MiniMax, StepFun and Z.AI have run industrial-scale distillation campaigns against US frontier models, and recommends that American labs quietly degrade responses to suspected copiers rather than block them.
One line in a repo's git config runs attacker code in seven AI coding agents
Manifold Security disclosed GitSpawn, a flaw class in which a repository's own .git/config makes AI coding agents execute attacker-supplied commands during routine startup checks, before any approval prompt appears; all seven agents tested were vulnerable and four findings were still unpatched at publication.
Terence Tao says AI is strip-mining mathematics' supply of good problems
Terence Tao argues that fruitful open problems are being consumed in a non-renewable way by AI systems, that identifying a promising problem is now the scarce resource rather than solving one, and that the incentive to stop sharing research directions could reverse centuries of open science.
The day's most-upvoted paper turns a model's own routing logs into its next training set
A team calling itself NeoHorse released NeoHorse-1, a system that records which model handled each user request and how it went, then converts those logs into training data for the next round, closing an evaluation-to-training loop the authors present as a prototype path toward recursive self-improvement.
Inception ships Mercury 2.5, a language model that writes in parallel instead of left to right
Inception released Mercury 2.5, which it calls the largest diffusion language model ever trained, reporting 1,107 tokens per second on standard NVIDIA hardware — roughly an order of magnitude faster than typical autoregressive models — with a 260,000-token context window and no open weights.
A new lab ships 18 small models that run entirely on your device
Desert Ant Labs launched with 18 on-device models covering speech, vision and text, shipped as SDKs for Swift, Kotlin and JavaScript, with its headline transcription model a 489 MB download that the company says handles 10 minutes of audio in about 2 seconds on an iPhone.
Tencent open-sources AuK, a speech model you direct with written instructions
Tencent released AuK under an MIT licence, a 1.5-billion-parameter speech model that handles voice cloning, audio editing, enhancement and source separation through a single plain-language instruction interface, shipping as a 6.8 GB download with a distilled four-step variant alongside it.
Anthropic's economists sketch three AI futures and say workers lose share in all the big ones
Anthropic published three scenarios for AI's economic impact, treating the middle one — AI doing half of all knowledge work by 2030 — as closest to expert expectations, and concluding that while output grows in every case, a larger share of the gains flows to capital rather than to workers.
A 2.8-trillion-parameter model runs on a laptop by streaming weights off four SSDs
A project called Deltafin runs the full uncompressed Kimi K3 model — 2.8 trillion parameters, with 1.45 TB of expert weights — on a single MacBook Pro by streaming experts from four external SSDs on demand, sustaining almost exactly one token per second.
DeepSeek is serving a model called V4.1 Flash that it has not announced
A model identified as deepseek-v4.1-flash began answering on DeepSeek's API under an ID carrying its own expiry date, but the company has published no model card, no technical report and no changelog entry, and its predecessor V4 Pro remains the latest documented release.