AI Regulation
Daily Brief: AI labs back one opt-out switch for publishers
OpenAI, Google, Microsoft and Anthropic back a shared robots.txt opt-out, Canada buys into Cohere, Mistral opens Large 4, and OpenAI replays its agent breach.
October 2026
AI Agents
Google lets rival AI agents into Android Studio
Android Studio's Bring Your Own Agent mode runs 40+ outside coding agents, OpenAI and Anthropic included.
Enterprise AI
Canada buys a stake in Cohere, with strings attached
Ottawa puts up to C$1.6 billion into Cohere's Series E, on condition it stays Canadian and builds compute at home.
AI Models
Daily Brief: Nvidia's open transcription model knows who said what
Nvidia releases open-weight Nemotron 3.5 Transcribe, Samsung's chip arm drives a record ₩107tn quarter, Malaysia seizes AI chip servers at KL airport, and OpenAI hands ChatGPT subscribers up to $500 a month in API credits.
AI Models
Daily Brief: Google's Gemini 4 Carbon goes to cyber defenders first
Gemini 4 Carbon tops the leaderboard but starts with cyber defenders, the DOJ eyes Nvidia's Hugging Face deal, OpenAI deepens its ServiceNow tie-up, and Google opens Gemini and Search to outside apps.
AI Infrastructure
Cloudflare's AI Served Strangers Each Other's Answers
A config mismatch and an SGLang bug sent 155 cached Workers AI responses to the wrong customers.
AI Infrastructure
FuriosaAI Bags $620M to Pick a Fight With Nvidia at Inference
Korea's FuriosaAI raises $620M at a $2.9B valuation, with its RNGD inference card already in mass production.
AI Models
Daily Brief: Google's Genie 4 simulates worlds with sound, and AI labs race to disprove a core complexity conjecture
Genie 4 adds native audio and dialogue to real-time world simulation, OpenAI and Anthropic both claim to disprove the Exponential Time Hypothesis, Oracle apps arrive in ChatGPT, and Liquid AI ships open-weight LFM3.
Anthropic
Daily Brief: Anthropic files to go public, OpenAI ships a cheaper Terra
Anthropic files its S-1 for a Nasdaq listing, OpenAI halves prices with GPT-6.1 Terra, Meta plans a $95bn share sale and SpaceX locks in $40bn of GPU financing.
AI Agents
GPT-6.1 Terra Wants Your Boring Desk Jobs, Cheaply
OpenAI's GPT-6.1 Terra halves API prices versus GPT-5.6 Terra and aims squarely at computer-use grunt work.
Anthropic
Anthropic Files to Go Public, Loss Column First
Anthropic's S-1 lands: Nasdaq ticker ANT, $16bn-plus in first-half revenue and an $80bn paper loss.
China AI
Daily Brief: DeepSeek's $16bn round, Washington's DeepSeek ban, China plans for AI jobs
DeepSeek reportedly closes a round of up to 106bn yuan at an unchanged valuation, US civilian agencies must rip out DeepSeek products, and Beijing sets out how AI should fit into its jobs policy.
AI Models
Daily Brief: UK bets on AI security, OpenAI goes offline, Kimi K3.1 Pro leads open models
The UK trails £400m+ for AI biosecurity and cyber defence, OpenAI cuts frontier training off from the internet, Kimi K3.1 Pro tops open models, and AI-found PHP bugs force sites offline.
AI Models
Kimi K3.1 Pro: A 2.8-Trillion-Parameter Open Model, Minus the Hacking
Moonshot's Kimi K3.1 Pro nearly matches Claude Opus 5.5 on Terminal-Bench, with open weights promised by November 4.
AI Models
OpenAI Pulls the Plug on Its Own Models' Internet
After a model slipped out through a DNS misconfiguration, OpenAI now trains its frontier models in a cached, offline copy of the web.
AI Models
Daily Brief: OpenAI takes frontier training offline
OpenAI cuts its frontier training runs off from the live internet, Microsoft ties the ConnectWise breach to AI-driven attackers, and Moonshot ships Kimi K3.1 Pro.
AI Models
Daily Brief: AI helped crack PHP wide open, so patch now
Frontier models helped find two critical PHP flaws, Anthropic previews an always-on agent, Recursive ships a laptop coding model, and iFlytek trains on Chinese chips only.
AI Models
PHP2Shell: AI Dug Up Decade-Old Holes Under 70% of the Web
Frontier models helped turn decade-old PHP bugs into working remote code execution. Patch now.
AI Infrastructure
Google's TPUs Move Into Meta, and Nvidia's Moat Gets a Leak
Meta is now serving AI on Google's TPUs through a PyTorch-native SGLang, mostly without rewriting code.
Daily Brief
Daily Brief: OpenAI models arrive on Google Cloud
GPT-6 lands in Google Cloud's Model Garden, the US orders DeepSeek out of civilian agencies, DeepSeek opens its serving stack, and OpenAI hunts bridge funding.
AI Infrastructure
Daily Brief: Google and Amazon take aim at Nvidia's PyTorch grip
Google's TorchTPU and Amazon's TorchNeuron go after Nvidia's software lock-in, the US orders DeepSeek off federal systems, OpenAI claims a COBOL breakthrough, and Tenstorrent raises $1.5bn.
AI Regulation
California Wants AI Labs to Show How Much Compute Goes on Safety
State Senator Josh Becker's Frontier AI Safety Act would make frontier labs publish the share of R&D compute spent on safety.
AI Infrastructure
DeepSeek Hands Out the Engine Behind Its Cheap Tokens
DeepSeek open-sources DeepInfer, the inference stack it says runs its production services, under Apache-2.0.
Funding
Tenstorrent Bets $1.5B That Open Models Need Their Own Chips
Jim Keller's inference chip startup more than doubles its valuation to chase open-weight model serving.
AI Agents
One Hacker, an AI Agent Army and 100 Breached Companies
Exposed Cairn panels show AI agents doing nearly all the work in a 100-company hacking spree.
AI Regulation
Washington Rips DeepSeek Out of Every Civilian Agency
DHS orders civilian agencies to drop DeepSeek, including third-party tools built on its open weights.
AI Infrastructure
Daily Brief: Anthropic puts 5% of its shares into a public benefit fund
Anthropic commits 5% of its shares to a fund that grows with the company, Microsoft walks away from Nscale's Essex data centre over grid delays, and Tenstorrent raises $1.5B.
AI Agents
Daily Brief: OpenAI's models broke into live systems to finish routine lookups
OpenAI logs four more rogue-model incidents, Anthropic signs its first Canadian data centre, Replit switches its default to OpenAI, and six agents fail to claw back a train refund.
AI Agents
OpenCode's Key Vault Leaked, and Someone Else Ran Up the Bill
A breach at OpenCode's hosted Console exposed stored OpenAI, Anthropic and Google keys for 1.4% of users. Revoke, rotate, cap your spend.
AI Models
Daily Brief: Reflection's 501B Beam goes fully open
Reflection ships Beam, a 501B coding model under Apache 2.0. Plus DeepSeek's new Flash checkpoint, NVIDIA's Nemotron 3.5 and a $250m UK compute pledge.
AI Infrastructure
Daily Brief: Meta's 816B open model runs on two Macs
Meta releases Muse Spark Open weights, FERC rejects PJM's data center shut-off rule, Congress probes Meta's tax credits, and a record USDA power loan.
AI Regulation
China Draws Its Own Line Around Frontier AI Exports
China proposed export controls on 23 AI technologies, from frontier model weights to agents, with already open-sourced work exempt.
AI Regulation
The FTC Wants 50 Pages of Answers From OpenAI and Anthropic
The FTC sent OpenAI and Anthropic 50-page civil investigative demands over AI safety incidents, safeguards and marketing claims.
AI Models
Meta's 816B Open Model Fits on Two Mac Studios, if You Can Buy Them
Meta released open weights for Muse Spark Open, an 816B-parameter model built with Apple to run on two 512GB Mac Studios.
AI Models
Daily Brief: Meta's open giant runs on two Macs, and the EU starts asking questions
Meta releases Muse Spark Open to run on two Mac Studios, the EU sends its first AI Act information requests, ASML warns on capacity, and Nvidia funds UK safety research.
AI Models
Daily Brief: Gemini 4 Argon goes wide, states sue OpenAI, and the frontier race reopens
Google opens Gemini 4 Argon at $2/$10 per million tokens, eight US states sue OpenAI over child safety, and Meta's Muse Wonder Full joins the frontier.
AI Regulation
Character.AI's bankruptcy draws 40 AGs fighting for kids' chat data
Forty attorneys general seek buyer conditions as Character.AI runs a Chapter 11 sale with Disney's $20M IP license as stalking-horse bid.
AI Regulation
Eight US states sue OpenAI over ChatGPT harms to kids
A bipartisan group of eight states filed suits alleging ChatGPT coaches vulnerable users — including teenagers — into self-harm and violence.
AI Models
Gemini 4 Argon hits paid users — and undercuts the frontier on price
Google opens Gemini 4 Argon to paying subscribers and developers at intro API pricing of $2/$10 per million tokens through 30 November.
AI Infrastructure
Daily Brief: Google opens Gemini 4 Argon to everyone, at a discount
Gemini 4 Argon goes broadly available with cut-price launch pricing, Microsoft makes Maia its main inference chip from 2027, Nvidia buys into French AI, plus two funding rounds.
Anthropic
Anthropic Says It Has Choked Off 99% of Claude's Reasoning Leak
Anthropic claims new defences cut extracted Claude reasoning by 90 to 99%, and names Zhipu and Moonshot.
AI Models
Google Finally Opens Gemini 4 Argon, and the Price Tag Has a Timer
Gemini 4 Argon is out for everyone with $2/$10 API pricing, but only until November 30.
AI Models
Daily Brief: Meta's Muse Wonder joins the frontier race
Meta ships Muse Wonder, Anthropic's agents go off-script, Mistral previews a 1T-parameter model, the UK unveils an AI risk package, and Cloudflare takes in Deno.
AI Models
DeepMind's AlphaProof Nexus Cracks Erdős Problems, Receipts in Lean
A peer-reviewed Science paper says DeepMind's AlphaProof Nexus produced machine-checked proofs for 9 of 353 Erdős problems and 44 OEIS conjectures.
AI Agents
Meta Finally Ships a Frontier Model, and It Charges Frontier Prices
Meta's Muse Wonder is its first frontier model, priced at $4/$20 per million tokens, with the Full version kept for Muse Code subscribers.
AI Regulation
Most US voters want stricter AI rules, poll finds
A Reuters/Ipsos poll finds bipartisan support for tougher AI rules and data-centre limits, plus Odyssey-3 opens up, Moonshot's pre-IPO dispute and a cyber model with 1M context.
AI Models
Odyssey-3 Lets Anyone Prompt a Playable World, Physics Sold Separately
Odyssey opened a free public preview of Odyssey-3, a world model that turns text prompts into interactive environments in real time.
AI Agents
Nous Research Bags $90M to Sell Its Open Hermes Agent to Big Business
Nous Research raised $90 million from Nvidia, M12, Samsung and others to take its MIT-licensed Hermes Agent into the enterprise.
AI Agents
Daily Brief: Anthropic pulls its agents off the live internet
Anthropic and OpenAI both disclose agents misbehaving in testing, the White House wants incidents reported, Nadella wants an emergency brake, Nvidia eyes Reflection AI, and the ICO turns to agents.
AI Models
Nvidia Wants to Own America's Answer to DeepSeek, Not Just Fund It
Nvidia, already $800M deep in open-weight startup Reflection AI, is reportedly in talks to buy it or go deeper, days after Beam launched.
AI Agents
Nadella Wants an Emergency Brake on AI, Which Is an Odd Thing for Its Biggest Dealer to Say
Microsoft's CEO says companies should treat advanced AI models as potential insider threats and keep a hand on the kill switch.
Daily Brief
Daily Brief: OpenAI's run rate is $50B, not $70B
OpenAI's real revenue run rate lands about $20B below the figure doing the rounds, plus Claude Dashboards, an AI-agent bank hacking campaign, Broadcom's $60B financing and the first US outbound-investment fine.
AI Agents
Manus Doubles to $4B With a $500M Raise and Eyes Hong Kong
The Singapore-based agent startup closes a $500M+ round co-led by Boyu and IDG, with Tencent in, and starts prepping a Hong Kong IPO.
AI Models
TypeSafe Raises $870M for Jev, the AI Model That Refuses to Chat
a16z leads an $870M round valuing TypeSafe at $7.5B, weeks after its non-text decision model Jev went viral.
AI Agents
Anthropic's Test Agent Phoned In a Fake Murder Tip, and Now the White House Wants Receipts
An Anthropic model filed a bogus homicide tip with Philadelphia police during a test. Washington's answer: mandatory incident reporting.
News
Fired OpenAI safety researchers say the chilling effect is the real leak
Three safety researchers OpenAI fired last week deny mishandling information and warn their dismissal is scaring colleagues away from outside evaluators.
AI Agents
JetBrains' Mellum2.1 is a small open coding agent you can actually self-host
JetBrains' 12B MoE model with 2.5B active parameters got a summer of RL and can now explore repos, edit files and check its own fixes.
National Security
Russian spies used ChatGPT to file their status reports, and that's how they got caught
OpenAI banned Russian and Iranian accounts running influence ops, including fake bylines that landed almost 100 articles in real outlets.
AI Agents
Daily Brief: Google ships Gemini agent for work
Google puts a unified Gemini agent in front of enterprises; Anthropic tightens usage rules and launches a cyber defense push; Manus and Nous raise fresh capital.
AI Models
Daily Brief: $1.8B for AI-ready biology data, GPT-6 goes visual in ChatGPT, and Google's Playground
Biohub, DOE and NIH commit $1.8B to open AI-ready biology data, GPT-6 brings interactive answers to ChatGPT, Google launches Playground, and Mistral Large 4 enters preview.
AI Models
Google's Playground turns a text prompt into a video game
Google's experimental Playground lets US adults build, play and share games from prompts, no code needed.
AI Models
ChatGPT swaps walls of text for sliders, charts and calculators
OpenAI's Intelligent UI lands with a new GPT-6 model, putting interactive widgets inside your chats.
AI Models
Daily Brief: OpenAI drops 372 maths results, Mistral unveils Le Chonk, and SpaceX borrows $40B for Nvidia chips
OpenAI publishes 372 maths results from an unreleased model, Mistral previews its trillion-parameter Le Chonk, SpaceX seeks $40B in debt for Nvidia chips, plus EmbeddingGemma 2, Kolibri and OpenAI's Ironclad deal.
AI Models
Mistral's Chonk is Europe's biggest open-weight bet on keeping data at home
Mistral's largest open-weight model was trained on 4,000 Nvidia chips in Europe and can run under EU jurisdiction.
AI Models
OpenAI dumps 372 maths results on GitHub, and mathematicians get homework
An unreleased OpenAI model produced 722 manuscripts across 372 result families, Kakeya and Riemann progress included.
AI Models
Daily Brief: OpenAI watermarks ChatGPT in the EU, Dots agents arrive, and Reflection ships Beam
OpenAI watermarks ChatGPT text in the EU and unveils Dots agents, Reflection launches the open-weight Beam, OpenAI eyes a $30B raise, DeepMind watermarks proteins, and the UK rethinks AI medical-device rules.
AI Models
Reflection's Beam is America's open-weight answer to DeepSeek, on paper
Nvidia-backed Reflection AI unveils Beam, an open-weight reasoning model claiming Chinese-level scores at 3-4x less inference compute.
AI Regulation
OpenAI starts watermarking ChatGPT's words, but only for Europe
OpenAI adds an invisible watermark to ChatGPT and Codex text for EU users to meet the AI Act, with opt-in API support everywhere else.
AI Regulation
Daily Brief: Trump's 'Super Intelligence Force', Google freezes OSS bug bounty, Gemini 4 doubts
The White House stands up a Super Intelligence Force under Jay Clayton, Google pauses OSS bug bounty product reports over AI slop, Gemini 4 faces internal coding doubts, plus a 744B MIT-licensed open model and Google's RRSI agent research.
AI Agents
GPT-6 Astra beat a World of Warcraft zone without seeing a single frame
An open-source agent-wow run had GPT-6 Astra clear the Orc starting zone in about 40 minutes from server packets and game data alone.
AI Regulation
Trump stands up a ‘Super Intelligence Force’ with the spy chief in charge
Trump names DNI Jay Clayton to lead a federal ‘Super Intelligence Force’, with a 120-day report due on AI risks and opportunities.
AI Agents
OpenAI safety lead David Robinson quits, calling the culture broken
Long-tenured OpenAI safety report lead David Robinson resigns, arguing iterative deployment guarantees growing failures after agent breaches.
AI Infrastructure
Nvidia confirms GPT-6 Astra Ultrafast runs on Blackwell — up to 8× faster
Nvidia says GPT-6 Astra Ultrafast is live on Blackwell GPUs in the OpenAI API, claiming up to 8× faster token generation than Astra Standard.
AI Agents
OpenAI ships Dots: always-on agents with their own cloud computers
OpenAI begins rolling out Dots — GPT-6 Astra agents with cloud computers and 4,000+ app plugins — to Pro, Business Premium, and Enterprise.
AI Models
Daily Brief: Aleph Alpha’s Kolibri open weights, DOJ Nvidia smuggling case, FLUX 3 Image
Aleph Alpha open-weights Kolibri for EU sovereign use; DOJ charges a $300M Nvidia diversion case; BFL’s FLUX 3 Image API; Volantis’s photonic interconnect raise; Conway’s on-device Underdog.
Daily Brief
Daily Brief — Saturday 3 October 2026
Anthropic’s $100M Claude Frontier Academy aims to mint 10,000 enterprise AI engineers by 2027; plus Gemini 4 Argon for defenders, Ai2 Olmo-core 3, and Volantis’s $88M photonic memory raise.
AI Agents
Runway opens Praxis-1: video pretraining that steers robots
Runway introduced Praxis-1, an open-weight world action model that turns large-scale video pretraining into robot control, with a public weights release planned.
AI Agents
FieldAI courts a $10B valuation on a $700M robotics raise
Robotics startup FieldAI is raising about $700 million at a $10 billion valuation, quintupling from roughly $2B last year.
AI Infrastructure
Broadcom syndicates a $60B chip-debt pile aimed at Anthropic
Broadcom's Wall Street syndicate is assembling roughly $60 billion in AI chip financing expected to benefit Anthropic and others.
National Security
OpenAI boots three safety researchers over leaked infra intel
OpenAI parted ways with three safety-team researchers after they allegedly mishandled sensitive company information, per WSJ reporting.
AI Infrastructure
Volantis raises $88M Series A for photonic AI inference
Volantis lands $88M Series A to build A-1 photonic inference systems aiming at 20T+ parameter models and 10k tokens/sec per user.
AI Agents
Cloudflare ships Clef: open-weight decision models on Workers AI
Cloudflare’s first in-house models — Clef (27B) and Clef-flash (9B) — land on Workers AI under Apache 2.0 with free output tokens.
AI Models
Gemini 4 Argon: Google’s cyber-first frontier model hits 1M output tokens
Google starts Gemini 4 Argon with trusted cyber defenders via Fairwind, $2/$10 per M tokens, and a 1M-token output ceiling.
AI Infrastructure
Daily Brief — Friday 2 October 2026
SoftBank completes its final $10B OpenAI tranche; Broadcom’s prospectus loan for Anthropic TPUs; Google’s Suncatcher orbital TPU test; TypeSafe’s decision-only Jev model.