Welcome to

Bytes

The day’s tech news, a few lines per story.

Scroll down to begin

BreakingZero-Days & Patching

QNAP Warns NAS Owners Over Active Plex2Shell CVE-2026-96652 RCE Attacks

  • QNAP issued an urgent advisory in November 2026 after attackers compromised NAS units running unpatched Plex Media Server.
  • Plex2Shell chains CVE-2026-96652, CVE-2026-96654, CVE-2026-96655, and CVE-2026-96656 for unauthenticated RCE over port 32400.
  • CISA added the flaws to its KEV catalogue as automated bots deploy cryptocurrency miners and proxy payloads.
  • Administrators must update to Plex Media Server 1.43.3 (build 1.43.3.10861) or later and audit exposed systems.

Zero-Days & Patching

Google Issues Emergency Chrome Patch for Actively Exploited XSLT Sandbox Escape CVE-2026-125350

  • Google issued an emergency patch for CVE-2026-125350, a high-severity use-after-free flaw in the browser XSLT component.
  • Threat actors actively chained the memory corruption vulnerability to achieve arbitrary code execution inside the renderer sandbox.
  • The US Cybersecurity and Infrastructure Security Agency added the flaw to its Known Exploited Vulnerabilities catalogue with a strict federal deadline.
  • Enterprise administrators and users must update Chrome to version 131.0.6778.85 or higher across Windows, macOS, Linux, and Android.

Semiconductors

Nvidia Robotics Engineer Darrell Dai Resigns Over How Jensen Huang's Exit Was Handled

  • Darrell Dai, a robot research engineer at Nvidia, said on X on 15 November that he is leaving the company.
  • He said staff were asked to accept Jensen Huang's departure before receiving an explanation, and that engineers had not felt heard on the decision.
  • Dai said he does not have all the facts about the allegations around Huang and would not prejudge the investigation.
  • His post passed 2.2 million views within hours, making it the most visible staff reaction since Colette Kress became interim chief executive.

Regulation & Trade

Italy Fines Meta €805 Million Over Default Android Photo Gallery Access

  • Italy's Garante fined Meta €805 million ($887 million), its largest penalty ever against a technology firm and 0.47% of Meta's global annual revenue.
  • The ruling targets Facebook's Android gallery suggestions feature active since 2023, which affected an estimated 20 million Italian users.
  • Regulators enforced Article 122 of Italy's Privacy Code under the ePrivacy Directive, bypassing Meta's Irish GDPR lead authority.
  • Meta has 60 days to disable the feature by default and build an explicit consent prompt, though the company plans to appeal.

Chips & Compute

Nvidia Approaches TSMC for Custom 3D DRAM Integration on Next-Generation AI Silicon

  • Nvidia has approached TSMC to explore direct vertical stacking of 3D DRAM layers onto next-generation AI accelerator logic tiles.
  • The architectural proposal seeks to bypass High Bandwidth Memory substrate bottlenecks by bonding memory dies directly using sub-micron hybrid copper bonds.
  • TSMC reiterated its operational commitment to pure-play foundry logic, confirming it will not enter commodity memory cell fabrication.
  • Pilot engineering evaluations between Nvidia, TSMC, and memory partners are slated for advanced packaging facilities in Tainan during late 2027.

Zero-Days & Patching

Critical GPUStack RCE Vulnerability CVE-2026-102895 Exposes Enterprise AI Compute Clusters

  • Cybersecurity researchers have identified CVE-2026-102895, a maximum-severity unauthenticated RCE flaw in open-source AI platform GPUStack.
  • The vulnerability allows remote attackers to bypass API authentication, execute arbitrary shell commands, and seize host GPU acceleration hardware.
  • Incident responders have observed threat actors hijacking clusters to siphon proprietary foundation model weights and deploy cryptomining tools.
  • GPUStack has released emergency version 1.2.4, urging all enterprise administrators to update immediately and isolate management ports.

Models & Research

OpenAI Dissolves Standalone Robotics Unit to Fold Hardware Work Into Core Models

  • OpenAI dissolved its standalone robotics research division and reassigned researchers to its core foundation-model and application teams.
  • Leadership concluded that general-purpose multimodal models can drive physical actions directly without an isolated robotics silo.
  • The move aligns OpenAI's architecture with unified vision-language-action pipelines rather than bespoke kinematics stacks.
  • Hardware partners and rival labs such as Google DeepMind and Xilect continue to invest heavily in embodied AI.

Zero-Days & Patching

CISA Orders Agencies to Patch Actively Exploited UnrealIRCd Flaw CVE-2026-114713 by 16 November

  • CISA added UnrealIRCd CVE-2026-114713 (CVSS 8.6) to the Known Exploited Vulnerabilities catalogue on 13 November with a 16 November deadline.
  • The command injection bug affects UnrealIRCd 6.0.0 through 6.2.7.0 in networks with two or more linked servers via the labelled-response module.
  • An unprivileged user can inject server-to-server commands such as GLINE and KILL with full server authority.
  • Admins must upgrade to UnrealIRCd 6.2.7.1, apply hot-patch bundle-20261107, or deploy the vendor spamfilter regex rule.

Agents & Tools

Google Launches Gemini Decisions API for 150ms Multimodal Routing

  • Google released the Gemini Decisions API powered by gemini-decision-1 in Google AI Studio and the Gemini API.
  • A single request accepts up to 262K multimodal input tokens and evaluates up to 1,000 independent questions.
  • The model emits zero output tokens, returning structured routing and rubric results in 150 to 250 milliseconds.
  • Pricing is set at $0.10 per 1 million input tokens with no output token charges.

Agents & Tools

vLLM vs SGLang vs TensorRT-LLM (2026 Comparison): Throughput, KV Caching, and Security

  • vLLM offers the broadest hardware compatibility (NVIDIA, AMD ROCm, TPU) and PagedAttention memory management under an Apache 2.0 licence.
  • SGLang leads on multi-turn agent and structured-JSON workloads via RadixAttention prefix caching, though recent RCE disclosures highlight strict IPC isolation needs.
  • TensorRT-LLM delivers peak NVIDIA Hopper and Blackwell throughput via ahead-of-time engine compilation, at the cost of slower model iteration.
  • Teams serving multi-tenant agentic workloads should pair prefix caching with network-isolated worker sockets and non-root containers.

Agents & Tools

Model Context Protocol Transition to Agentic AI Foundation Formalises Enterprise Tool Calling Standard

  • The Model Context Protocol transferred governance to the Linux Foundation Agentic AI Foundation as a neutral open standard.
  • Founding contributors Anthropic, OpenAI, and Google DeepMind established unified schema specifications for local and remote MCP servers.
  • The standard eliminates brittle point-to-point API adapters by providing secure JSON-RPC interfaces for resources, prompts, and tools.
  • Enterprise implementations gain standardised authentication, credential isolation, and audit logging across autonomous developer agent loops.

Agents & Tools

OpenAI and Card Networks Launch ChatGPT Wallet for Autonomous Agentic Payments

  • OpenAI has unveiled ChatGPT Wallet in partnership with Visa, Mastercard, and American Express.
  • The protocol issues dynamic, cryptographically bounded digital payment tokens with hard user-defined spending caps and merchant allowlists.
  • Autonomous AI agents can execute flight bookings, software licence purchases, and supply reorders without exposing primary card numbers.
  • Rollout begins for ChatGPT Enterprise and Team subscribers before expanding to third-party developers via the OpenAI Payments API.

Models & Research

Google DeepMind AlphaProof Code Formally Verifies Signal Protocol Machine Code at AArch64 Level

  • DeepMind researchers have applied AlphaProof Code to formally verify the cryptographic core of the Signal messaging protocol.
  • The system generated machine-checked Lean 4 mathematical proofs certifying memory safety and timing-channel resistance down to AArch64 binary instructions.
  • The automated verification pipeline eliminated manual proof translation by linking high-level protocol specifications directly to compiled object code.
  • The verified codebase and proof certificates have been submitted to the Signal Foundation for integration into production libraries.

Models & Research

Google DeepMind Unveils AlphaGenome 2 with Single-Cell Resolution for Genetic Medicine

  • Google DeepMind has launched AlphaGenome 2 to model human genomic sequences at single-cell resolution.
  • The neural architecture predicts chromatin accessibility across 1,200 distinct cellular states directly from raw DNA sequences.
  • Clinical validation against UK Biobank cohorts demonstrated a 34% accuracy gain in identifying non-coding pathogenic regulatory variants.
  • The weights and inference API are available to academic researchers under open-access biological research licences.

Asia & SEA Tech

MAS Opens Future of Finance Institute With PathFin.ai and SafeFin.ai

  • MAS officially began operating the Future of Finance Institute on 3 November 2026, led by interim CEO Alan Lim.
  • The institute organises its AI rollout around three pillars: PathFin.ai, BuildFin.ai, and SafeFin.ai.
  • BuildFin.ai aligns with MAS's Safeguards for Agentic Finance at Runtime (SAFR) framework for autonomous financial agents.
  • SafeFin.ai expands the MindForge AI Risk Management Toolkit for third-party model testing and assurance.

Asia & SEA Tech

MAS and SSFA Ease Data Centre Power Limits for Water-Saving Facilities Across Asia

  • The SSFA and Monetary Authority of Singapore launched the second edition of the Singapore-Asia Taxonomy at COP31 in Antalya.
  • New formulas let data centres outside Singapore trade slightly higher PUE for lower water consumption in water-stressed Asian regions.
  • Green classification requires PUE <= 1.40 - (0.0545 x WUE) with WUE capped at 2.2, while existing transition facilities face a 1.45 - (0.04545 x WUE) limit.
  • Facilities inside Singapore remain governed by the stricter BCA-IMDA Green Mark standards.

Asia & SEA Tech

Singapore Replaces PSG with S$30,000 EDGE Grant as IMDA Targets Enterprise AI Automation

  • Enterprise Singapore and IMDA phased out legacy Productivity Solutions Grant applications in favour of the streamlined EDGE Grant framework.
  • Qualifying enterprises can secure up to S$30,000 in direct co-funding to deploy artificial intelligence and deep-tech workflow automation.
  • Pre-approved digital solution categories prioritise autonomous customer operations, secure internal data synthesis, and ERP telemetry.
  • The policy shift aligns with Singapore National AI Strategy 2.0 to transition regional businesses from basic digitisation to agentic infrastructure.

Funding & Deals

DayOne Secures $320M Round to Scale Enterprise Context Engines Across Asia-Pacific

  • Enterprise context infrastructure startup DayOne has completed a $320 million funding round at a $2.4 billion post-money valuation.
  • The capital will accelerate expansion of real-time knowledge graph engines for sovereign banking clients in Singapore, Tokyo, and Sydney.
  • The platform allows enterprises to anchor large language models to internal data stores without exfiltrating sensitive records to external cloud APIs.
  • Net revenue retention across regional financial enterprise deployments reached 142% over the trailing twelve months.

Funding & Deals

Staera Semiconductor Raises Pre-Seed Capital to Commercialise On-Chip LED Optical Interconnects

  • Singapore-based Staera Semiconductor has closed pre-seed financing led by regional deep-tech venture syndicates.
  • Spun out of Singapore's A*STAR research institutes, the startup develops monolithic micro-LED optical interconnects for advanced chip packages.
  • The optical links achieve 100 times higher bandwidth density than conventional copper wire bonding while reducing inter-chip power draw by 65%.
  • Initial silicon test vehicles are scheduled for packaging validation with commercial foundry partners during the second quarter of 2027.

Asia & SEA Tech

Singapore Captures 92% of Southeast Asian Venture Capital as Late-Stage Mega-Rounds Mask Seed Slump

  • Singapore captured 92% of total venture capital deployed across Southeast Asia during the latest reporting quarter.
  • Late-stage mega-rounds into artificial intelligence infrastructure and data centre developers masked an underlying 28% drop in seed deal volume.
  • Regional venture firms have tightened due diligence, requiring Series A applicants to demonstrate unit economics and positive gross margins.
  • Institutional capital from North America and the Middle East continues to concentrate within Singapore-domiciled corporate structures.

Growth & IPOs

Anthropic Inks $28B 15-Year Data Centre Compute Deal With Riot Platforms Ahead of Nasdaq IPO

  • Anthropic has finalised a 15-year commercial lease agreement with digital infrastructure developer Riot Platforms to secure up to 910 megawatts of high-density powered land at a dedicated campus in Corsicana, Texas.
  • The commercial contract, detailed in updated regulatory filings ahead of Anthropic's expected initial public offering on the Nasdaq, represents approximately $28 billion in baseline commitments over the initial 15-year term.
  • The transaction underscores the structural shift across the artificial intelligence sector, where foundation model laboratories are directly securing utility interconnection rights and substation capacity to guarantee compute availability…

Growth & IPOs

Anthropic Launches Nasdaq IPO Roadshow Under Ticker ANT at $70 to $78 Per Share

  • Anthropic, PBC officially kicked off its initial public offering roadshow on November 10, moving one of the world's largest private artificial intelligence laboratories onto public markets under the proposed Nasdaq Global Select Market…
  • According to Anthropic's roadshow filing announcement, following its October 26 S-1 filing with the US Securities and Exchange Commission, the proposed offering totals 825,000,000 Class A common shares priced in an indicative range of…

Enterprise & Cloud

Microsoft 365 21-Hour Outage Exposes Single-Provider Enterprise Risk

  • A global Microsoft 365 disruption ran roughly 18 to 21 hours from 12 November into 14 November 2026.
  • Exchange Online, Outlook, Teams, SharePoint, OneDrive, Copilot, and Defender XDR suffered timeouts and queued mail.
  • Microsoft ruled out a cyberattack, blaming a server configuration and network capacity change that saturated connection handling.
  • Disruptions across hospitals, schools, and airlines revived calls for multi-vendor resilience and digital sovereignty.

Enterprise & Cloud

Terraform vs OpenTofu vs Pulumi in 2026: Licensing, State, and Multi-Cloud Costs

  • OpenTofu holds an open-source MPL-2.0 licence under the Linux Foundation, while Terraform uses HashiCorp's Business Source Licence (BSL).
  • OpenTofu ships native client-side state encryption, whereas Terraform relies on backend storage encryption or HCP Terraform.
  • Pulumi lets engineering teams write infrastructure in TypeScript, Python, Go, or C# instead of HCL.
  • Team pricing diverges at scale: self-hosted OpenTofu carries zero seat fees, while HCP Terraform and Pulumi Cloud charge per-resource or per-credit.

Big Tech & Platforms

Google Workspace to Disable Search for K-12 Schools That Block Gemini AI

  • Google Workspace for Education will disable Gemini tools and Google Search by default for K-12 student accounts starting in early 2027.
  • School districts that keep Gemini disabled will also lose student access to Google Search because AI Overviews are tied to Search.
  • US lawmakers and educators including Senator Bernie Sanders and former FTC chair Lina Khan criticised the bundling as coercive.
  • Administrators in districts such as New York City and Los Angeles must decide whether to opt in or deploy alternative search engines.

Enterprise & Cloud

Self-Hosted AI Agents Comparison 2026: OpenClaw vs Goose vs Agent Zero Architecture and Security

  • OpenClaw delivers messaging-first agent orchestration across WhatsApp, Telegram, and Slack with local credential persistence.
  • Block's Apache-licensed Goose operates directly inside developer command lines with MCP server extensibility and minimal overhead.
  • Agent Zero embeds dynamic subagent spawning inside isolated Docker containers, offering superior security boundaries for untrusted execution.
  • Enterprise deployment choices depend on IPC latency requirements, multi-tenant credential segregation, and hardware compute overhead.

Big Tech & Platforms

Meta Freezes Updated Commercial Data Terms Indefinitely Following Advertiser and Regulatory Backlash

  • Meta has suspended planned updates to its commercial advertising terms following pushback from brand advertisers and data protection regulators.
  • The contested terms would have granted Meta broad rights to use advertiser creative assets, product catalogues, and CRM data to train foundation models.
  • Major global brand coalitions threatened to pause digital advertising expenditure across Facebook, Instagram, and WhatsApp.
  • European data protection authorities warned the changes could breach GDPR consent rules governing commercial audience data.

Enterprise & Cloud

Data-Eng-Bench Establishes Open-Source DBA and Data Engineering Evaluation Standard for AI Agents

  • dbt Labs, Snowflake, and university researchers have launched Data-Eng-Bench to evaluate AI agents on end-to-end data engineering and DBA operations.
  • The benchmark moves beyond simple text-to-SQL by assessing model performance on live dbt pipeline compilation, schema migrations, and warehouse cost tuning.
  • Frontier models achieved only a 41% success rate on multi-step schema refactoring tasks, identifying frequent hallucinations in dependency macros.
  • The evaluation suite runs entirely inside isolated Docker sandboxes with reproducible Postgres, Snowflake, and BigQuery execution targets.

Breaches & Threats

SignalWire FreeSWITCH SIP Command Injection CVE-2026-116779 Targeted in Carrier Telephony Probes

  • Cybersecurity researchers have detected automated exploitation probes targeting CVE-2026-116779 in SignalWire FreeSWITCH VoIP servers.
  • The vulnerability allows remote attackers to execute arbitrary shell commands by injecting unescaped metadata into Session Initiation Protocol headers.
  • Compromised PBX systems have been abused to intercept carrier audio streams, conduct toll fraud, and route unauthorized international telephony traffic.
  • SignalWire has released maintenance releases for FreeSWITCH branches 1.10 and 2.0, urging carriers to apply patches immediately.

Zero-Days & Patching

CISA Adds OpenSSH SSHole Authentication Bypass CVE-2026-123158 to Known Exploited Vulnerabilities Catalogue

  • The US Cybersecurity and Infrastructure Security Agency has added CVE-2026-123158 to its Known Exploited Vulnerabilities catalogue.
  • The vulnerability, dubbed SSHole, stems from an unexpected status return code in OpenSSH daemon builds on Linux systems.
  • Unauthenticated remote threat actors have conducted mass IPv4 scans to trigger state confusion and bypass key exchange authentication.
  • Federal civilian agencies must remediate vulnerable internet-facing OpenSSH servers by November 28 under Binding Operational Directive 26-04.

Zero-Days & Patching

Apache Discloses Critical mod_http2 Heap Corruption Flaw Enabling Remote Code Execution

  • The Apache Software Foundation has issued an emergency security release for the Apache HTTP Server, resolving a critical pre-authentication heap memory corruption vulnerability in the core HTTP/2 protocol implementation module.
  • The security flaw, tracked as CVE-2026-119618, affects Apache HTTP Server releases from version 2.4.55 through 2.4.70.
  • The maintainers have addressed the flaw in Apache HTTP Server version 2.4.71, recommending that system administrators across Linux distributions and enterprise web tiers upgrade immediately.

Zero-Days & Patching

CISA Issues Emergency Patch Mandate for Check Point Gaia OS Flaw Chained With Wiper Attacks

  • The United States Cybersecurity and Infrastructure Security Agency has added a critical local privilege escalation vulnerability affecting Check Point network security appliances to its Known Exploited Vulnerabilities catalogue, ordering…
  • The flaw, tracked as CVE-2026-123039, carries an initial Common Vulnerability Scoring System rating of 8.8.
  • According to threat intelligence advisories from incident response teams, threat actors are actively chaining the privilege escalation vulnerability with secondary initial access vectors to deploy destructive wiper malware inside…

Breaches & Threats

OpenAI Rates GPT-6 Cyber Critical and Activates 35-Second Control Room Kill Switch

  • OpenAI has formally classified its newest offensive-and-defensive security model, GPT-6 Cyber, at the "Critical" tier under its internal Preparedness Framework, prompting the company to withhold general consumer access and ship a…
  • The dual rollout, announced across technical disclosures between 5 November and 9 November 2026, marks the first time OpenAI has paired a frontier security model release with a mandatory, fail-closed inference kill switch designed to…

Robotics & Defence

Intel Pushes Core Ultra Series 3 Into Humanoid and Commercial Robotics

  • Intel is positioning its 18A Core Ultra Series 3 (Panther Lake) processors for commercial service and humanoid robots.
  • Integrating CPU, GPU, and NPU on one SoC cuts battery drain, heat, and bill-of-materials cost versus discrete GPUs.
  • Deployments include Sensory AI's Ella autonomous barista kiosk and Oversonic Robotics' RoBee humanoid.
  • Intel is extending Series 3 from AI PCs into certified industrial and embedded edge automation.

Chips & Compute

Infineon Produces First Functional 300mm Silicon Carbide Power Wafers in Villach

  • Infineon fabricated functional 300mm (12-inch) silicon carbide power devices at its Villach, Austria plant and showed the wafer at electronica 2026.
  • Moving from 200mm to 300mm SiC wafers produces roughly 2.3 times more chips per wafer using adapted silicon fabrication equipment.
  • At its Dresden site, Infineon also demonstrated 300mm Cold Split wafering to reduce crystal slicing waste.
  • Commercial volume ramp schedules will be set with automotive, renewable energy, and AI data centre power customers.

Chips & Compute

TSMC and Amkor Finalise Advanced CoWoS Packaging Alliance to Relieve AI Silicon Bottlenecks

  • TSMC and OSAT provider Amkor Technology finalised a strategic back-end manufacturing alliance spanning facilities in Arizona and Gwangju.
  • The partnership distributes advanced packaging stages for Chip-on-Wafer-on-Substrate technology to alleviate global assembly constraints.
  • Foundry packaging demand expanded more than twofold year-over-year, driven by Nvidia Rubin and AMD Venice server accelerator orders.
  • TSMC implemented packaging fee adjustments of up to 15% as high-performance computing customers bid for guaranteed substrate capacity.

Robotics & Defence

Google DeepMind Unveils Gemini Robotics 2.5 Vision-Language-Action Architecture for Dexterous Manipulation

  • Google DeepMind has introduced Gemini Robotics 2.5, a unified vision-language-action foundation model for robotic control.
  • The architecture bridges multimodal spatial reasoning with a 200-hertz real-time diffusion motor trajectory policy for dexterous robotic hands.
  • In benchmark manipulation trials, the model achieved an 88% success rate on zero-shot deformable object handling and precision assembly tasks.
  • The system runs locally on edge accelerator modules, allowing industrial robots to operate without continuous cloud connectivity.

Chips & Compute

Brown University Physicists Eliminate AI Chip Dead Layer to Slash Neural Accelerator Power Draw

  • Physicists at Brown University have resolved the dead-layer interface defect that previously crippled atomic-thin multiferroic memory materials.
  • The material breakthrough allows magnetoelectric logic gates to switch non-volatile magnetic states using pure electric fields rather than electrical currents.
  • Experimental prototype cells demonstrated a 90% reduction in switching energy compared to standard complementary metal-oxide-semiconductor logic.
  • The research, published in Nature Materials, provides a clear physical roadmap for building zero-standby-power neuromorphic inference processors.

Chips & Compute

Qualcomm Implements 10% Price Increase on PC Processors as Advanced Foundry Costs Escalate

  • Qualcomm has confirmed a 10% price increase across its Snapdragon X-series PC processor portfolio starting in early 2027.
  • The pricing adjustment reflects higher contract wafer fabrication and advanced packaging costs charged by TSMC at leading-edge nodes.
  • Original equipment manufacturers, including Dell, Lenovo, and HP, are assessing consumer retail price adjustments for next-generation AI PCs.
  • Qualcomm reported strong adoption in enterprise fleet upgrades, citing energy efficiency advantages over legacy x86 architectures.

Space & Frontier

SpaceX Catches Returning Starship Upper Stage Mid-Air on Flight 15

  • SpaceX Flight 15 caught the returning Starship upper stage (Ship 42) mid-air using the Mechazilla tower chopstick arms at Starbase on 14 November.
  • Ship 42 spent more than 22 hours in low Earth orbit across roughly 14 orbits and deployed 26 Starlink V3 satellites before re-entry.
  • This is the first time an orbital-class rocket upper stage has returned from space and been caught at its launch site.
  • Super Heavy Booster 22 aborted its return-to-pad manoeuvre due to a guidance and propellant issue and splashed down offshore in the Gulf of Mexico.

Biotech & Health

Beam Therapeutics Files Historic FDA Application for Base Editing Therapy Following Phase 3 Trial

  • Beam Therapeutics submitted a Biologics Licence Application to the US FDA for risto-cel to treat severe sickle cell disease.
  • Risto-cel represents the first human therapeutic application of base editing, modifying target DNA letters without generating double-strand breaks.
  • Phase 3 trial telemetry showed durable non-expression of sickling haemoglobin and complete resolution of severe vaso-occlusive crises in 96% of cohort patients.
  • The FDA granted priority review status under its accelerated cellular and gene therapy platform evaluation pathway.

Space & Frontier

Rocket Lab Reaches 100th Electron Launch Milestone with Orbital Deployment of iQPS Radar Satellite

  • Rocket Lab has completed its 100th commercial orbital mission with the successful launch of an Electron rocket from Launch Complex 1 in Mahia, New Zealand.
  • Electron becomes the fastest commercially developed launch vehicle in history to reach 100 flights, achieving the milestone in eight years.
  • The mission, designated 'The Mountain God Rises', deployed the YAMATSUMI-II synthetic aperture radar satellite for Japanese commercial operator iQPS.
  • The flight demonstrated rapid launch cadence and ocean stage recovery as Rocket Lab prepares to debut its medium-lift Neutron rocket.

Climate & Energy

COP31 Global Carbon Budget Projects First 0.4% Dip in Fossil Emissions Led by Solar and EV Additions

  • The Global Carbon Project released its 2026 report at COP31, projecting a 0.4% decline in global fossil fuel carbon emissions to 38.2 billion metric tons.
  • Record renewable energy deployments, led by solar photovoltaic additions and electric vehicle adoption, offset growing global electricity demand.
  • Emissions in Europe and the United States continued their structural decline, while emissions in major emerging economies approached their structural plateau.
  • Scientists cautioned that total emissions remain dangerously elevated due to ongoing deforestation and extreme climate-driven land degradation.

Climate & Energy

Hyperscalers Accelerate SMR Nuclear Deals to Meet 24/7 Clean Energy Demands for AI Clusters

  • Global cloud hyperscalers are committing billions of dollars in venture equity and long-term power purchase commitments to advanced nuclear fission startups, accelerating the commercial development of Small Modular Reactors (SMRs) to power…
  • Over the past six months, Amazon Web Services, Google, Microsoft, and Meta Platforms have announced commercial partnerships with advanced nuclear developers, including Kairos Power, Oklo, X-energy, and Constellation Energy.
  • The corporate investments reflect an urgent operational reality: the rapid expansion of gigawatt-scale AI data centres is outstripping available electrical grid capacity, while corporate net-zero climate commitments prevent technology…

Climate & Energy

Meta Signs Long-Term Clean Cement, Near-Zero Steel, and Freight Deals for Data Centres

  • Meta on November 12 disclosed a broad package of multi-year purchase commitments across low-carbon cement, coal-free primary steel, electric heavy freight, and chemically recycled plastics tied directly to its North American AI data centre…
  • According to Meta Newsroom, long-term corporate offtake agreements are structured to give first-of-a-kind industrial facilities predictable revenue to finance commercial plants.

Fintech & Commerce

UK FCA Launches Quantum Experimentation Group for AML and Fraud

  • The UK FCA announced the Quantum Experimentation Group (QEG) in November 2026, calling it the first regulator-led quantum finance group.
  • Participants will benchmark Quantum Machine Learning against classical models for anti-money laundering and fraud detection.
  • The FCA is supplying a synthetic transaction dataset to avoid exposing confidential customer records.
  • Expressions of interest close on 31 December 2026 ahead of a Q1 2027 programme launch.

Regulation & Trade

US Voters Reject 98% of Local Data Centre Ballot Measures Over Grid Strain

  • Voters rejected or restricted roughly 98% of local data centre ballot measures across 10 US states in November 2026.
  • Ohio hosted more than 20 measures around Columbus, while states from California to Wisconsin passed bans or 25 MW caps.
  • Residents cited rising household electricity bills, grid brownout risk, and closed-loop cooling water demand.
  • Contrast with Singapore's MAS/SSFA taxonomy shows operators now face strict power and water gates globally.

Regulation & Trade

Canadian Court of Appeal Rejects OpenAI Motion to Dismiss News Publisher Copyright Class Action

  • The Court of Appeal for Ontario has dismissed an appeal filed by OpenAI, allowing a sweeping copyright infringement class action brought by a consortium of major Canadian news publishers to proceed to trial.
  • The litigation, initiated by media organisations including The Globe and Mail, Torstar Corporation, Postmedia Network, The Canadian Press, and CBC/Radio-Canada, alleges that OpenAI systematically scraped and reproduced millions of…
  • The appellate court's unanimous decision upholds a lower court ruling that dismissed OpenAI's jurisdictional challenge and request for summary dismissal, establishing that Canadian courts possess subject-matter jurisdiction to hear claims…

Fintech & Commerce

Monetary Authority of Singapore Unveils SAFR Framework for Autonomous Agentic Financial Execution

  • The Monetary Authority of Singapore has published a supervisory framework governing the deployment of autonomous artificial intelligence agents within regulated financial institutions, establishing runtime safety boundaries,…
  • The guidance document, titled Safeguards for Agentic Finance at Runtime (SAFR), was developed under the central bank's BuildFin.ai initiative in collaboration with commercial banking institutions, technology infrastructure providers, and…
  • The release arrives ahead of the opening of the Singapore FinTech Festival 2026 at the Singapore EXPO, where global central bankers and fintech executives will convene to address the systemic implications of autonomous financial software…

You’re up to date

New stories land here through the day. The full archive is on the front page.

Go to the front page