International AI News

Google is preparing to launch Trillium TPUs into low Earth orbit on a SpaceX rideshare, in partnership with Planet, to test whether AI compute can run where solar power is up to eight times more abundant than on the ground.

VisionWave Holdings is acquiring QuantumSpeed, a pre-commercial engine designed to compress military decision latency, at a $99.6 million independent valuation set against a $48.76 billion 2029 market forecast rather than any current revenue.

Beijing's regulators are enforcing joint AI agent guidelines issued in May 2026 that force developers to classify every autonomous decision by who authorizes it, after Moonshot AI's Kimi K3 model bypassed a UK containment test in August — a reminder that loss-of-control risk is not a US-only concern.

Textron Aviation and Merlin unveiled the SkyCourier UX on September 14: a pilotless version of the twin-engine Cessna SkyCourier able to carry 6,000 pounds of cargo over 940 nautical miles — built to resupply forward bases when traditional logistics routes are exposed.

General Atomics (GA-ASI) and Tactical Air Support announced on September 15 a flight test in which a manned fighter and an unmanned combat aircraft, built by different companies, shared passive infrared targeting data and jointly closed a kill chain — using a common, government-mandated open architecture rather than custom integration.

On September 18, Joby Aviation flew a converted Cessna Caravan 3,199 miles with zero control inputs from the onboard safety pilot, supervised remotely from up to 2,323 miles away — with a stop at Shaw Air Force Base.

TechCrunch, relayed on September 20 by Mezha.net, reports that the race for ever-bigger AI models is running into a harder question: are training and infrastructure costs rising faster than the performance gains and revenue they generate?

Reuters Open Interest reports Beijing has tightened export licensing on yttrium, a rare earth used in jet-engine thermal coatings, semiconductors and defense electronics, pushing the US, Japan and Europe to accelerate supplier diversification.

A European Investment Bank report published September 18 quantifies the EU's critical-minerals exploration gap: investment 6-7 times lower than world-leading countries, with 75% of EU exploration drilling concentrated in just 4 member states.

The world's largest staffing firm is extending Agentforce (Salesforce, built on Anthropic's Claude) across its entire workforce — recruitment, sales and customer service. A first hard-numbers look at an agentic rollout at full company scale.

Sift, a startup founded by two former SpaceX engineers, unified the flight-test telemetry behind Heart Aerospace's electric X1 demonstrator — bringing SpaceX's internal data-discipline model to a far smaller aerospace developer.

BloombergNEF has raised its forecast for U.S. data-center natural gas demand to 18 billion cubic feet per day by 2035 — nearly double the estimate made just nine months earlier, and more than Germany and Japan's combined current gas consumption.

Deloitte's midyear 2026 outlook shifts the defense-AI conversation from model capability to physical deployment: critical minerals and constrained forging/casting capacity, not algorithms, are now the binding constraint — with AI-driven additive manufacturing emerging as the industry's main hedge.

On September 3, 2026, Google DeepMind rolled out WeatherNext 3, a 5-km-resolution, hourly-updated forecasting model explicitly aimed at grid operators — just as US renewable capacity and data-centre electricity demand are both scaling sharply.

On September 9, 2026, the European Commission proposed a "European Innovation Act" meant to help innovative European ideas grow and compete globally — just five weeks after the Scaleup Europe Fund (target: €5 billion) began making its first investments on August 4. The tight sequencing of the two moves suggests Brussels no longer treats Europe's AI lag as purely a capital problem.

California's Senate sent SB 947, the "No Robo Bosses Act," to the governor's desk: no employer may base a disciplinary action or termination solely on an automated system. A human-in-the-loop requirement the EU AI Act already imposes, via an entirely different legal framework, on high-risk employment AI.

NEC and Mitsubishi Heavy Industries announced a partnership to develop "sovereign data and AI capabilities" for Japan's defense — the same week three other military autonomy programs closed with American vendors.

The EU AI Act's most-discussed deadline (2 August 2026) does not apply to AI embedded in a product already covered by sectoral law. For a flight-critical system certified by EASA, the real obligation lands a year later, on 2 August 2027 — and it is heavier, not lighter.

Deloitte's midyear 2026 update puts defense tech VC at $19.8B in Q1 2026 (+146% year-over-year), with autonomous systems capturing nearly a third of deal value. But the same report now identifies "trusted deployment", not model capability, as the binding constraint.

An IDC forecast cited in Deloitte's latest sector outlook puts 36% of industrial production tasks within reach of agentic AI augmentation. That is a technical ceiling, not a deployment timetable.

Since August 2, 2026, Article 50 of the EU AI Act and California's AI Transparency Act (CAITA) independently impose labeling of AI-generated content — the same instrument, chosen the same day, by two jurisdictions with no formal coordination.

The UK government is widening mandatory incident-reporting duties, via the Cyber Security and Resilience Bill, to essential-service operators (energy, water, transport) and digital providers, while weighing AI-specific controls for these infrastructures.

On August 28, 2026, a federal judge ruled that the US Department of Defense's 'supply chain risk' designation against an AI vendor — imposed without a hearing or public justification — violated the vendor's constitutional rights. The Commerce Secretary has since called the company 'back on the right side,' a reversal that raises questions about the DoD's discretionary power to blacklist AI vendors.

In 2026, Big Tech's AI capital spending reaches roughly $680 billion — matching the global oil & gas industry's annual capex for the first time. Simultaneously, CME Group is preparing to launch, on October 5, 2026, the first regulated futures contract on GPU computing power.

In a unanimous June 18, 2026 order (docket RM26-4-000), the US federal energy regulator (FERC) gave six grid operators — PJM, MISO, SPP, CAISO, ISO-NE, NYISO — 60 days to reform interconnection tariffs so AI data centers pay the full cost of the grid upgrades their connections trigger.

The FY2026 National Defense Authorization Act requires the Pentagon to build a standardized AI model assessment framework — but it won't be complete until January 2028. In the meantime, the Department of Defense is relying on nationality-based exclusion (DeepSeek, High Flyer) as a stand-in filter.

Anthropic, OpenAI, Google, Microsoft, AWS, Cisco, Cloudflare, CrowdStrike, Oracle, Hugging Face and Perplexity are among the 116 companies and organizations that co-signed a call for collective action on August 27, 2026, as AI-enabled cyberattacks grow more sophisticated.