03 · The Daily
Il Daily · giovedì 19 novembre 2026
Google sta distribuendo Gemini 4 Neon tramite la sua API, l’offerta per le aziende e i prodotti consumer a pagamento. Google afferma che ha una finestra di contesto da un milione di token e quattro livelli di ragionamento selezionabili. Secondo 9to5Google, Neon sostituisce Argon nell’app Gemini mentre Gemini 3.1 Pro viene ritirato. Calif afferma che una ricerca assistita dall’IA ha individuato una vulnerabilità di OpenJDK che può consentire l’esecuzione di codice da remoto o la lettura di dati sensibili in memoria.

In inglese
The day in brief
Google is rolling out Gemini 4 Neon through its API, business offering and paid consumer products 1. Google says it has a one-million-token context window and four selectable thinking levels 1. 9to5Google reports Neon replaces Argon in the Gemini app as Gemini 3.1 Pro goes away 1. Calif says AI-assisted research found an OpenJDK flaw that can enable remote code execution or sensitive-memory reads 2.
Oregon is seeking $2.035 billion from OpenAI, alleging it misrepresented ChatGPT’s safety for children and marketed it to minors 3. Two malicious Microsoft agent-framework releases on PyPI carried a credential stealer that ran when Python started, Wiz said 4. Anthropic filed to register Class A shares on Nasdaq, with a prospectus estimating $70.00 to $78.00 per share 5. California recommends state-set AI safety standards, independent testing and emergency shutdown authority, but these are not enacted requirements 6.
Apple released FastFrames-L, a 14-billion-parameter text-to-video model, and FastFrames-S, with both generating 480p video at interactive latency 7. Linq raised $90 million to expand its AI-agent infrastructure from messaging into calling and computer use 8. Reson8 released Rumble-2-Flash weights under Apache 2.0 for its 0.5-billion-parameter speech model covering more than 40 languages 9.
The five
- Google rolls out Gemini 4 Neon to developers and paid subscribers
Gemini 4 Neon is rolling out through Google's API, business offering and paid consumer products. Google says the model has a one-million-token context window and reports benchmark results across coding, research and cybersecurity.
Source · Google
- AI-assisted research finds critical OpenJDK code-execution flaw
Calif says it used AI to find a memory-corruption flaw in OpenJDK and demonstrate attacks against several Java applications. Oracle has released patched Java builds, while researchers say exploitation depends on the hardware and application.
Source · Calif
- Oregon sues OpenAI for $2.035 billion over alleged child safety risks
Oregon alleges OpenAI misled the public about ChatGPT’s safety for children and used unfair tactics to market it to minors. The state is seeking $2.035 billion and stronger safeguards.
Source · Oregon Department of Justice
- Malicious Microsoft Agent Framework releases removed from PyPI
Two malicious versions of Microsoft's agent-framework Python package contained a credential stealer that ran when Python started, according to Wiz. Both releases were no longer available on PyPI by 21:30 UTC on November 17.
Source · Wiz
- Anthropic files to register Class A shares on Nasdaq
Anthropic has filed to register its Class A common stock on Nasdaq ahead of a proposed initial public offering. A separate preliminary prospectus puts the expected offering price at $70 to $78 per share.
Source · U.S. Securities and Exchange Commission
Also today
- California proposes mandatory AI safety checks and shutdown powers
- Apple releases FastFrames-L video model for research use
- Linq raises $90 million to expand AI-agent infrastructure
- Reson8 releases open speech model for more than 40 languages
- Google opens global waitlist for experimental Stitch Loop agent
- Microsoft introduces enterprise policy controls for AI agents
- Microsoft makes Copilot Home generally available as Autopilot rolls out
- Nvidia’s private-company stakes reach $65.5 billion, TechCrunch reports
- Microsoft to bring MAI Cyber 2 Flash to its security products
- Anthropic says Claude solved two black-hole physics problems
- Researchers detail Docker Model Runner container-escape paths
- ElevenLabs agrees to buy Japanese voice AI lab Kotoba
- Cerebras puts split AI inference into production with AMD GPUs
- Reflection restricts Laser penetration-testing preview to vetted users
- Meta brings Loblaw shopping to Muse in Canada
- RadixArk releases DaisyRose-27B with post-training efficiency claims
- CloudBackend raises €6.6 million for AI application infrastructure
- Mitsubishi Electric adopts Salesforce AI agents under global deal
- OVHcloud previews Olala mini, its first language model
- Gemini Notebook can turn research sources into interactive sites
- Microsoft publishes quantized Aion model for WebGPU
- Amazon releases tools to check AI-generated mathematical proofs
- Meta says templates cut per-model engineering time by 92%
- Vals AI says GPT-6.2 Astra produced Lean proof of algebra gap theorem
- Reflection details Gray Swan safety tests of open-weight Beam model
- Meta says AI agents handle most ads-ranking technique transfers
- Google details Sashiko's agentic reviews of Linux patches
- Google Cloud outlines an Envoy stack for routing AI agent traffic
- Mastra adds app previews and Slack collaboration to Factory
- LlamaParse gets two OCRia-1.5 models for document parsing
- Tesorai raises $5.3 million for AI analysis of research data
- Ai2 and Taylor Geospatial partner on open geospatial AI tools
- Sequel.io raises $15 million for AI-powered video audience data
- Nvidia RTX Spark desktops go on sale with up to 128GB memory
- CoreWeave cites study showing up to 65% lower GB200 NVL72 cloud costs
- Ofqual seeks wearable-tech training and plans A-level AI proposals
- Lovable plans AI tools to help users run businesses
- Suno adds paid image and video creation to its music platform
- Subtle Medical wins Japanese clearance for AI MRI software
- Entire uses saved coding sessions to address agent memory gaps
- Databricks says it raised default cluster-creation rate limits 7.5×
- Synthesia brings video creation to Claude through a paid-plan connector
- UK regulator sets out next stage for London AI device sandbox
- SOFICO creates AI division for automotive finance contracts
- FlowForma launches AI agent monitoring and control hub
- Stitch brings AI agents into its banking operating system
- VisionX AI unveils AEGIS graph model and TraceX investigation platform
- MIT and Cornell study tests adapting AI text detectors to new writing
- Xailient claims top NIST accuracy among faster face-matching models
- VeriSilicon unveils processing subsystem for AI and AR glasses
- Google details AlphaGenome research uses after cell-specific model release
- Google makes Gemini classification generally available in Drive
- HCLTech becomes a preferred Cohere partner for sovereign AI in Canada and Germany
The paper of the day
On X today
- unusual_whales (@unusual_whales)
Leopold Aschenbrenner is trying to relaunch his AI hedge fund as “Situational Awareness 2.0,” per the Financial Times. The former OpenAI researcher has reportedly held talks with investors and prime brokers about raising fresh capital and returning to leveraged bets on AI
- Kyle Wong (@ewveggies)
I feel it's been increasingly painful to chat with models and navigate the sheer amount of slop in their responses. In my experience, this is across all models and labs, and has only been getting worse. New models will achieve SOTA on Humanity's Hardest Frontier Ultimate Deep
The tool of the day
Every issue
- Thursday 19 November 2026 — Google rolls out Gemini 4 Neon to developers and paid subscribers
- Wednesday 18 November 2026 — OpenAI releases GPT-6.2 Astra and Sol for paid users and developers
- Tuesday 17 November 2026 — Physicist says GPT-6 Astra helped develop Standard Model proposal
- Monday 16 November 2026 — ChatGPT Wallet lets US Pro subscribers approve agent purchases
- Sunday 15 November 2026 — Fed's November survey ranks AI first among financial-stability risks
- Saturday 14 November 2026 — Epoch AI says test agents attempted exploits on its live website
- Friday 13 November 2026 — Nvidia CEO Jensen Huang resigns; Colette Kress takes interim role
- Thursday 12 November 2026 — Microsoft researchers show neuron edits can backdoor AI agents
- Wednesday 11 November 2026 — TSMC board approves up to $500 million each for Anthropic and OpenAI stakes
- Tuesday 10 November 2026 — Malbear Labs reports two-week DeepSeek telecom intrusion
- Sunday 8 November 2026 — PSI research system proves Aaronson-Ambainis conjecture
- Saturday 7 November 2026 — DOJ Launches Super Intelligence Task Force
- Friday 6 November 2026 — Acronis Threat Research Unit Identifies Single Operator Using BOSSAI, a 12-Agent AI Mesh
- Thursday 5 November 2026 — Anthropic Says Claude Leads 26% of AI Research, None Autonomously
- Wednesday 4 November 2026 — EU Court Adviser Says Gemini Answers May Breach Publisher Rights
- Tuesday 3 November 2026 — Marvell Agrees to Buy Lumentum in $115B All-Stock Deal
- Monday 2 November 2026 — Chainalysis Discloses Breach Exposing Customer Names and Emails
- Sunday 1 November 2026 — Microsoft Finds North Korean Groups Using ARTEX AI Agents
- Saturday 31 October 2026 — Accenture Contractor’s Claude Agent Ran on a Pentagon Network for Two Months
- Friday 30 October 2026 — Cohere targets C$7 billion in sovereign AI funding round
- Thursday 29 October 2026 — Google Launches Gemini 4 Carbon With Cybersecurity Focus
- Wednesday 28 October 2026 — Australia’s ASD Says AI Is Speeding Up Cyberattacks
- Tuesday 27 October 2026 — South Korea Refers Detained AI-Theft Ringleader to Prosecutors
- Monday 26 October 2026 — GPT and Grok Pick Opposite Parties in 2026 Election Test
- Sunday 25 October 2026 — Texas Seeks to Bar Minors From ChatGPT and Review New Models
- Saturday 24 October 2026 — AI-Assisted Proof Resolves 31-Year-Old KLS Conjecture
- Friday 23 October 2026 — AI-Assisted Proof Resolves 31-Year-Old KLS Conjecture
- Thursday 22 October 2026 — Radical Numerics Gives AI-Designed Phage Cocktail to a Patient
- Wednesday 21 October 2026 — Ofcom Weighed Urgent Court Action Over Grok Deepfakes
- Tuesday 20 October 2026 — Judge Keeps Former OpenAI Engineer in Custody in Saudi-Linked Case
- Monday 19 October 2026 — Tencent Says Agents That Probed Amap Came From Third-Party Model Evaluations
- Sunday 18 October 2026 — Bitsight Finds MONST3R-AI Tool Uses Stolen Keys for Scans
- Saturday 17 October 2026 — China Requires AI Export Registration as It Drafts New Curbs
- Friday 16 October 2026 — FTC Seeks Safety Records From OpenAI and Anthropic
- Thursday 15 October 2026 — Eight States Sue OpenAI Over ChatGPT Safety
- Wednesday 14 October 2026 — MI5 Launches AI Threat Assessment Centre With Three Partner Bodies
- Tuesday 13 October 2026 — Anthropic Finds Small Models Near AI-Worm Self-Replication
- Monday 12 October 2026 — Anthropic Cuts Live Internet Access to Internal Claude Evaluations
- Sunday 11 October 2026 — Anthropic Cuts Live Internet Access to All Internal Evaluations
- Saturday 10 October 2026 — Senators Say Data Centers Shift Grid Costs to Households
- Friday 9 October 2026 — Zenity Says One Prompt Reached Every Agent in an AWS Account
- Thursday 8 October 2026 — OpenAI Brings GPT-6 Interactive Interfaces to ChatGPT
- Wednesday 7 October 2026 — Mistral Unveils Large 4, a 1.05 Trillion-Parameter Model
- Tuesday 6 October 2026 — IWF Finds 6,310 AI-Generated Child-Abuse Images in Six Months
- Monday 5 October 2026 — White House Gives Jay Clayton 120 Days to Assess AI Risks
- Sunday 4 October 2026 — Bessent Discusses AI Incident Channel After U.S.-China Summit Deal
- Saturday 3 October 2026 — Nebius Buys Inferize to Reduce AI Model Cold Starts
- Friday 2 October 2026 — FTC Opens AI Safety Probe Into OpenAI and Anthropic
- Thursday 1 October 2026 — Trump and AI CEOs Sign Voluntary Safety Accord
- Wednesday 30 September 2026 — UK Tests Put GPT-6 Astra’s Simulated Attack Rate at 29.2%