China Mass‑Produces DUV as Hyperscalers Double Down on U.S. Compute

AI & Machine Learning

Researchers report training three‑dimensional visual foundation models directly on millions of routine CT and MRI scans substantially improved performance on downstream neuroimaging tasks and generalization across hospitals, showing scale plus clinical‑context training can materially raise utility in medical imaging. The Nature Medicine paper outlines how 3D pretraining yields robustness on varied clinical data and suggests practical dataset‑scale approaches for building medical foundation models while emphasizing the need for careful clinical validation before deployment. Authors discuss privacy, consent and dataset governance as key barriers to translation, and the work renews debate about how to move medical foundation models from research to regulated clinical settings. The study is likely to accelerate interest and investment in domain‑specific foundation models while prompting regulators and hospitals to tighten validation and data‑provenance requirements. Source: Nature Medicine Verified: True

Anthropic published a research post describing experiments using Claude to surface new approaches for analyzing and attacking certain cryptographic constructions, demonstrating both the analytical potential and the dual‑use risks of powerful LLMs in cryptanalysis. The post frames the work as exploratory research and emphasizes safety practices and responsible disclosure, but it also underscores how model access can accelerate discovery of vulnerabilities in domains historically dominated by specialist techniques. Anthropic calls for community engagement to weigh benefits against risks and to develop norms for publishing sensitive model‑driven findings. The release renews conversations about access controls, red‑teaming, and how to handle model outputs that may aid adversarial research. Source: Anthropic Research Verified: True

Consumer Hardware

Leaked marketing renders published by The Verge show Google’s upcoming Pixel 11 Pro Fold with larger foldable displays, revised hinge mechanics, and an updated camera module, offering an early look at the company’s premium foldable direction ahead of an official launch event. The renders — sourced from trusted leakers — suggest Google is aiming to close the hardware feature gap with other high‑end foldables while refining durability and pocketability. If accurate, the design changes signal Google’s commitment to a multi‑device hardware strategy that pairs larger, more flexible displays with tighter integration of software‑based camera and AI features. The leak sets expectations for Google’s product roadmap and competitive posture in the foldable market segment. Source: The Verge Verified: True

DoorDash announced “DoorDash Air,” a new drone delivery division launched after the FAA granted approvals for commercial drone operations, and plans to operate company‑owned drone routes for restaurant and grocery deliveries in limited markets initially. The move positions DoorDash as a direct operator in airborne last‑mile logistics rather than just a marketplace partner, signaling a strategic bet on autonomous delivery to reduce per‑delivery costs and delivery times. Early operations will be tightly scoped to approved corridors and markets, giving DoorDash an operational playbook to scale if regulations and economics prove favorable. The announcement intensifies competition among incumbents and startups pursuing drone logistics and will influence municipal and FAA considerations for broader commercial drone corridors. Source: The Verge Verified: True

Cybersecurity

Cyera disclosed a letter of intent to acquire Oasis Security for roughly $1 billion, a move the data‑security vendor says will boost protections for proliferating AI agents and cloud data flows as enterprises scale agentic workflows. The planned acquisition is pitched as embedding stronger data protection directly into AI pipelines and agent orchestration layers, addressing a growing enterprise need to control exfiltration and exposure from automated agents. If completed, the deal would reflect a rising market for specialized controls tailored to AI‑driven workloads and could accelerate consolidation among security firms targeting model‑centric risks. Observers will watch how Cyera integrates Oasis’s tooling and whether customers see immediate benefits in lineage, policy enforcement, and runtime monitoring. Source: TechCrunch Verified: True

Rapid7 published a detailed advisory on “KindaRails2Shell” (CVE‑2026‑66066), a critical Active Storage vulnerability in Ruby on Rails that enables arbitrary file read and may allow remote code execution under certain configurations. The Rapid7 analysis includes reproduction steps, enumerates impacted Rails versions and configurations, and provides mitigation guidance and detection suggestions for web operators. Rapid7 urges immediate patching and inventorying of affected systems because the flaw can expose sensitive application files and potentially allow escalation when combined with specific deployment setups. The advisory underscores the continued importance of rapid dependency management and timely framework updates in modern web‑app ecosystems. Source: Rapid7 Verified: True

Enterprise Infrastructure

Reuters reports Chinese firms have begun mass production of domestically built immersion DUV lithography machines, marking a significant step toward onshoring critical chipmaking equipment for mature‑node semiconductor production. The move could reduce reliance on Western suppliers for immersion DUV tools used in many foundry lines and help accelerate local capacity expansion for established process nodes. While the machines do not replace the most advanced EUV tools, wider availability of domestic DUV gear may ease bottlenecks for fabs producing chips for automotive, power management, and other mature‑node markets. The development will be watched by global supply‑chain planners and policymakers as part of broader efforts to localize semiconductor ecosystems. Source: Reuters Verified: True

Meta and BlackRock announced a roughly $14 billion joint venture to build and operate a major data‑center campus near El Paso, Texas, marrying Meta’s hyperscale compute demand with BlackRock’s real‑estate capital and signaling continued hyperscaler expansion in U.S. compute capacity. The project highlights how hyperscalers are leveraging third‑party capital partners to accelerate large campus builds while spreading financial risk, with material implications for local jobs, grid load and regional infrastructure planning. For communities and utilities near proposed campuses, these deals often trigger negotiations over power, transmission upgrades, and incentives, making the announcement immediately relevant to regional stakeholders. The scale of the investment also illustrates ongoing demand for coastal and inland compute capacity as AI and cloud workloads grow. Source: Reuters Verified: True

Commonwealth Fusion Systems announced an additional $1 billion financing to extend its runway for prototype milestones and industrial engineering efforts, reinforcing the firm’s strong cash position in the private fusion sector. The fresh capital is intended to accelerate prototype timelines and de‑risk next‑stage engineering work as CFS moves toward demonstration systems, reflecting investor appetite for capital‑intensive, long‑horizon energy technologies. While fusion commercialization remains a multi‑year challenge, the funding round signals continued belief among backers that scaled magnetic confinement approaches could reach industrial relevance. The raise will enable CFS to sustain development cadence and supplier commitments as it advances toward larger integration milestones. Source: TechCrunch Verified: True

Policy & Regulation

Politico reports EU policymakers are reframing AI safety enforcement to focus on rogue agents, model provenance, and geopolitically sensitive risks, proposing stricter mandates for provenance auditing and controls over agentic operations in response to recent incidents. The reporting outlines how the EU seeks a regulatory architecture that balances operational security, platform accountability, and strategic competition concerns — particularly in relation to China — while building enforcement tools that can act quickly on high‑risk agent deployments. The shift reflects lessons from incidents where autonomous agents produced unexpected or risky behaviors and signals a move from broad principles to operational rules that target deployment vectors. Stakeholders across industry and government will need to align on technical standards and compliance mechanisms as the EU operationalizes these enforcement aims. Source: Politico.eu Verified: True