Uber and Rivian Partner on Autonomous Ride-Hailing Fleet
Uber and Rivian have forged an exclusive partnership to deploy Rivian’s new R2 autonomous vehicles on the Uber platform. The deal includes a $300 million investment from Uber and outlines a multi-year rollout across major U.S. cities. This collaboration will leverage electric autonomous vehicles to bolster Uber’s ride-hailing services and expand its autonomous supply. The strategic move gives Uber more direct control over the hardware for its autonomous fleet, with plans to deploy vehicles in cities like San Francisco and Miami through 2031.
Sources:
Google Launches Gemma 4: A New Family of Open-Weight AI Models
Google has launched Gemma 4, its latest family of open-weight AI models, built on the same advanced technology as its Gemini 3 models. The Gemma 4 family is available in four sizes to cater to diverse applications, from mobile devices to data centers. The smaller ‘Effective’ models (2B and 4B parameters) are optimized for on-device tasks, while the larger 26B Mixture-of-Experts and 31B dense models are engineered for more demanding workloads. Google highlights that Gemma 4 delivers significant improvements in reasoning, code generation, and complex logic handling. These multimodal models can process text, images, and, for the smaller versions, audio. Gemma 4 is released under the Apache 2.0 license, permitting broad commercial use.
Sources:
Oracle Unveils Secure AI Data Platform for US Federal Agencies
Oracle has announced its AI Data Platform, a new offering tailored for U.S. federal agencies. This platform is designed to help government entities securely connect generative AI models with their proprietary data and applications. The primary goal is to dismantle information silos and accelerate informed decision-making for both civilian and defense missions. The platform integrates Oracle Cloud Infrastructure (OCI), Oracle Autonomous AI Database, and OCI Enterprise AI. It is engineered to meet the federal government’s stringent security and compliance standards, operating within OCI’s FedRAMP High-authorized Government Cloud.
Moonbounce Secures $12M to Launch AI Control Engine
Moonbounce, a startup developing an AI control engine, has emerged from stealth with $12 million in funding. The round was led by Amplify Partners and StepStone Group. Moonbounce’s platform is designed to provide organizations with real-time governance over their AI systems, ensuring they operate predictably and safely. The company’s patented control engine translates content policies into predictable AI behavior, aiming to overcome the limitations of traditional content moderation in the generative AI era. The platform is already being used by clients in sectors like dating apps and generative content websites.
Sources:
Autodesk Debuts Wonder 3D for Generative AI 3D Asset Creation
Autodesk has launched Wonder 3D, a new generative AI model now integrated into its Autodesk Flow Studio platform. This powerful tool allows creators to generate editable 3D assets from text prompts or images. Wonder 3D is designed to streamline 3D workflows by cutting the technical complexity and time needed to create 3D content. The model supports text-to-3D, image-to-3D, and text-to-image generation, making it accessible to both professionals and newcomers in game development, film, and marketing. The objective is to enable rapid prototyping and iteration while maintaining full creative control.
Sources:
Microsoft Pledges $10B for Japan’s AI and Cloud Infrastructure
Microsoft has announced a $10 billion investment in Japan from 2026 to 2029 to significantly expand its artificial intelligence and cloud computing infrastructure. This landmark deal represents Microsoft’s largest-ever single-country overseas investment and includes a focus on strengthening cybersecurity cooperation with the Japanese government. The investment involves partnerships with SoftBank and Sakura Internet to bolster Japan’s sovereign AI computing capabilities, enabling sensitive data to be processed domestically on Microsoft’s Azure platform. As part of this initiative, Microsoft aims to train one million engineers and developers in Japan by 2030, supporting the nation’s technological advancement and national security.
Sources:
Niobium Introduces ‘The Fog’ for Fully Encrypted AI Cloud Workloads
Niobium has unveiled ‘The Fog,’ a private cloud infrastructure platform that enables organizations to process AI and data workloads on information that remains fully encrypted, even during active computation. The platform leverages the power of fully homomorphic encryption (FHE), ensuring decryption keys remain exclusively with the data owner and that Niobium has zero access to customer data. This breakthrough addresses a major cloud computing challenge by allowing the use of sensitive and regulated data without exposure, a critical capability for industries like healthcare and finance. ‘The Fog’ is available in a private beta, with a public launch planned for Q2 2026. Niobium is also offering an SDK to help developers build applications for the platform without requiring deep encryption expertise.
Sources:
SharonAI Signs $1.25 Billion Deal for AI Cloud Infrastructure
SharonAI Holdings Inc. has secured a five-year, $1.25 billion agreement with ESDS Software Solutions Ltd. for AI infrastructure. Under the terms of the deal, SharonAI will deploy an 8K B300 cluster in an Australian data center. Revenue from this contract is projected to begin in the third quarter of 2026. The agreement also includes a customer option to extend for an additional two years. This contract marks a significant expansion of SharonAI’s AI Cloud services, and the company reports robust demand across enterprise, hyperscale, research, and government sectors.
Sources:
Kubernetes 1.36 Preview: Key Features Include Scale-to-Zero and Enhanced AI Support
The upcoming Kubernetes v1.36 release, anticipated on April 22, 2026, is set to introduce several pivotal features. The highly requested HPA Scale-to-Zero feature is expected to be enabled by default, allowing the Horizontal Pod Autoscaler to scale workloads down to zero replicas—a key feature for optimizing costs in non-production environments. For AI/ML workloads, Dynamic Resource Allocation (DRA) is being improved with support for taints and tolerations for AI hardware like GPUs and TPUs, offering more granular control over resource allocation. This release will also mark the retirement of the Ingress-Nginx controller, with the community recommending a migration to the more expressive and feature-rich Gateway API for managing external cluster access. Additionally, v1.36 will introduce manifest-based admission control configuration, ensuring security policies are active before the API server begins processing requests.
Sources:
Azure DevOps March 2026 Updates: Remote MCP Server and Enhanced Security
The March 2026 updates for Azure DevOps are now generally available, rolling out several new features and improvements. A key addition is the public preview of the Remote MCP Server, which enables AI-powered integrations without requiring a local server. Security is a key focus, with enhancements including new Advanced Security pull request status checks that integrate with branch policies, improved audit logging for secret push protection bypasses, and the ability to export security overview data to a CSV file. Azure Pipelines also sees improvements with better deployment visibility via artifact tracking and a new Stages view for a clearer overview of deployments across environments.
CavilinQ Raises $8.8M Seed Funding for Quantum Interconnects
Cambridge-based hardware startup CavilinQ has closed an $8.8 million seed funding round led by QVT, with participation from Safar Partners, MFV Partners, Serendipity Capital, and Harper Court Ventures. The company is developing critical interconnect hardware to link multiple quantum processors, aiming to address the physical scaling limitations inherent in single-processor designs. The new capital will fund a specialized laboratory in Cambridge, MA, and expand its engineering team to develop production-ready prototypes. CavilinQ’s technology uses cavity-enhanced photonic links to create high-fidelity light-matter interfaces, which will initially be integrated with neutral atom quantum processors.
Sources:
Microsoft Releases Open-Source Toolkit for AI Agent Security
Microsoft has introduced the Agent Governance Toolkit, an open-source project designed to provide runtime security for autonomous AI agents. The toolkit, released under an MIT license, is engineered to mitigate all ten agentic AI risks identified by OWASP, such as goal hijacking, tool misuse, and rogue agents. It provides a policy engine to intercept agent actions, secure agent-to-agent communication, and implement dynamic execution rings. The Agent Governance Toolkit is available for multiple programming languages, including Python, Rust, TypeScript, Go, and .NET. Microsoft’s long-term goal is to transition the project to a foundation for broader community governance.
Sources: