Tenstorrent Unveils RISC-V AI Workstation with Open-Source Stack
Tenstorrent announced the TT‑QuietBox 2, dubbed Blackhole, a liquid‑cooled AI workstation that runs directly on a desktop. The system can infer models up to 120 billion parameters and delivers teraflop‑class performance using a RISC‑V processor. It ships with a completely open‑source software stack covering compiler, runtime and kernel. Pricing starts at $9,999, making high‑end AI compute accessible at the desk.
Spain’s Sánchez Launches AI Tool to Track Hate Speech on Social Media
Spanish Prime Minister Pedro Sánchez unveiled HODIO, an AI‑driven system that ranks social‑media platforms by the volume of hate speech they host. The tool will scan publicly available posts, apply natural‑language processing, and publish a comparative score for each major...
Code Concepts: A Large-Scale Synthetic Dataset Generated From Programming Concept Seeds
Researchers introduced a concept‑driven workflow that produces synthetic code data aligned with specific programming skills. Using a taxonomy of 91 Python concepts, they generated roughly 15 million Python problems and incorporated 10 billion tokens into the final 100 billion‑token pretraining of Nemotron‑Nano‑v3. The...
Amazon Wants to Be Your First Stop for Health Questions —Here’s What Its New AI Assistant Does
Amazon has launched Health AI, an artificial‑intelligence assistant embedded in its website and mobile app that lets shoppers check symptoms, receive possible causes, and get treatment suggestions. The tool can direct users to telehealth visits or nearby providers, extending Amazon’s...
FreeWheel Launches AI Agent Infrastructure for Premium Video Ads, Pilots with PMG
FreeWheel unveiled an AI agent infrastructure built on its Model Context Protocol (MCP) Server and a suite of Intelligence tools, embedding artificial intelligence directly into premium video ad transactions. The platform enables real‑time workflow automation, performance monitoring, and decision‑making for...
OpenAI's New Training Dataset Teaches AI Models Which Instructions to Trust
OpenAI unveiled the IH‑Challenge dataset, a reinforcement‑learning resource that teaches models a four‑level instruction hierarchy—system, developer, user, and tool. The dataset replaces subjective LLM judges with deterministic Python scripts, enabling reliable automated evaluation. Early testing on the internal GPT‑5 Mini‑R...
Anthropic’s Pentagon Showdown Is Drawing Silicon Valley Into a Larger Fight
Anthropic has sued the Pentagon after the Department of Defense labeled the company a "supply chain risk," effectively blacklisting it from government contracts. A coalition of 37 leading AI researchers, including figures from Google, OpenAI and DeepMind, filed an amicus...
Takase Kanageta Selects Plataine to Scale Production Scheduling Across Multi-Plant Operations
Takase Kanagata, a Japanese precision‑manufacturing firm, has chosen Plataine’s AI‑powered Production Scheduler, delivered with Argosy Japan Aerospace Materials, to support its 2030 Digital Factory and Industry 4.0 roadmap. The solution will enable Takase to scale production scheduling across multiple plants while...
AI-Enabled Watch Towers Set to Proliferate Along the Border
The Department of Homeland Security will retrofit 148 uncrewed camera towers along the U.S. border with AI‑enabled sensors this year and add 50 next‑generation towers. GDIT’s Relocatable Autonomous Surveillance Tower brings longer‑range optics, radar, LIDAR and solar power, allowing edge...
Publish Your Data, AI Techniques, and Agentic Engineering Work on Towards Data Science
The New Stack is inviting its readers to publish on Towards Data Science, one of the largest AI‑focused publications. Submissions are free, and accepted pieces receive editorial support, rapid two‑day turnaround, and promotion across TDS’s channels. Authors can earn money...
Siemens to Help Build AI-Ready Scientific Infrastructure as Part of DOE’s Genesis Mission
Siemens announced a memorandum of understanding with the U.S. Department of Energy to support the Genesis Mission, a federal effort to modernize America’s scientific infrastructure with AI‑driven computing and interoperable digital systems. The partnership leverages Siemens’ expertise in industrial AI,...
TruScore Launches AI Capabilities for Its 360 Feedback Platform to Accelerate Leader Development
TruScore has added three AI‑powered features—AI Feedback Insights, AI Executive Summaries, and AI Comment Summaries—to its 360‑degree feedback platform. The tools automatically analyze quantitative scores and open‑ended comments, delivering concise themes, strengths, blind‑spot alerts, and development priorities. Built on a...
AI Podcast Start-Up Rebel Audio Comes Out of Stealth Mode, Taps Mark Burnett as Adviser (EXCLUSIVE)
Rebel Audio, a Nashville‑based AI podcasting start‑up, has emerged from stealth mode with a $3.8 million seed round and the addition of TV legend Mark Burnett as its first adviser. The company unveiled an invitation‑only beta of its AI‑driven platform at SXSW,...
The Latest Author Protest Against AI
Around 10,000 UK authors have jointly released an empty volume titled *Don’t Steal This Book* to protest the use of their works in training AI systems. The book, containing only a roster of contributors, will be distributed at the London...
Health IT Leaders Must Stay on Top of AI that Delves Into Clinical Workflows
At HIMSS 2026, Medicomp Systems CEO David Lareau warned that health systems are rapidly embedding generative AI into clinical documentation, coding and decision‑support workflows. While large language models promise efficiency, their probabilistic nature can introduce inaccuracies into patient records, jeopardizing...
AI Still Needs Consultants—For Now
OpenAI and Anthropic have signed agreements with major consulting firms to embed generative AI solutions across corporate clients, signaling a short‑term boost for the consulting sector. Analysts caution that while demand for AI‑focused advisory services will surge, the broader consulting...
Elsevier Launches Study Companion Osmosis AI to Transform How Medical Students Learn
Elsevier has launched Osmosis AI, a study companion that integrates its 140‑year medical content with the Osmosis video library and conversational AI. The platform delivers AI‑generated answers that are fully cited and paired with relevant videos, supporting USMLE Step 1 and...
An Efficient, Reusable Framework to Evaluate AI Safety
Researchers at Johns Hopkins and Microsoft introduced Jailbreak Distillation (JBDistill), a renewable framework that automates safety testing for large language models (LLMs). By over‑generating adversarial prompts and selecting the most effective subset, the method creates a consistent benchmark that can...
U.S. Government Growing Interest in Robotics
On March 10 the U.S. Department of Commerce hosted a roundtable with leaders from robotics, artificial intelligence, manufacturing and investment to chart how the United States can scale Physical AI‑enabled robotics. The discussion covered domestic supply‑chain strengthening, talent pipelines, digital...
Hidonix Is Now a Defense Company
Hidonix Industries announced a strategic shift from commercial applications to a primary focus on defense and public safety, leveraging its spatial intelligence, human‑centric AI, and robotics expertise. The company will target defense operators, emergency responders, and mission‑critical missions requiring precise...
Microsoft: AI Tools Now Handle 50M Health Questions Daily
Microsoft reported that its consumer AI tools, including Bing and Copilot, now field more than 50 million health‑related questions each day. An analysis of over 500,000 anonymized health conversations in January 2026 showed health is the top topic on mobile devices,...
Honeycomb Advances Observability for AI-Powered Software Development
Honeycomb announced a suite of AI‑enhanced observability features and the general availability of Honeycomb Metrics. New Agent Skills let Claude Code, Cursor and other AI agents migrate telemetry, create boards and set SLOs autonomously. Automated Investigations, a Slackbot with chain‑of‑thought...
Meta Is Developing 4 New Chips to Power Its AI and Recommendation Systems
Meta announced four new MTIA chips—300, 400, 450 and 500—to power its generative‑AI features and content‑ranking systems. The MTIA 300, already in production, will handle training for recommendation algorithms, while the 400‑500 series focus on inference workloads. Built on open‑source RISC‑V...
Jamdesk Launches First API Documentation Platform With Built-In AI Chat at No Extra Cost
Jamdesk, founded by former Ayrshare executives, launched a docs‑as‑code platform that embeds AI chat and natural‑language search in every plan. The service converts MDX files into live documentation sites, auto‑detects GitHub changes, and deploys to a global CDN in under...
HIMSS26: WVU Medicine Expands Abridge Ambient AI Platform Across 25 Hospitals
WVU Medicine announced expansion of Abridge’s ambient AI platform to over 2,800 clinicians across its 25‑hospital network. The rollout, driven by clinician advocacy, aims to alleviate physician burnout in rural West Virginia. Post‑implementation surveys of more than 200 providers reported...
Governing Perception
The piece argues that governance is the connective tissue that aligns trust with trustworthiness by managing perception across corporate, AI, data, and operational domains. It shows how gaps between how stakeholders perceive a system and how it actually performs drive...
Quizlet Launches as Native App in ChatGPT to Transform AI-Powered Learning
Quizlet has launched as a native app inside OpenAI’s ChatGPT, allowing users to generate flashcards and study sets directly from AI conversations. The integration taps into the platform’s library of hundreds of millions of study sets, merging generative AI with...
Fractal Unveils Intelligent Sales Agents to Accelerate B2B Growth
Fractal introduced Flyfish.ai, an AI‑native revenue acceleration platform that deploys more than 35 coordinated intelligent agents across the entire B2B sales lifecycle. Early enterprise pilots report up to 30% faster deal cycles and a 42% lift in sales productivity by...
Replit Introduces Agent 4 to Treat Software Development as Creative Work
Replit unveiled Agent 4, the latest iteration of its AI‑powered development platform, positioning software creation as a creative activity. The new system is structured around four foundational pillars that aim to bridge imagination with concrete implementation. While some fear AI could...
RAG Shows Its Work. That’s Not the Same as Being Right.
At the Generative AI Summit Austin, Ramkumar Shanker warned that the death of third‑party cookies forces publishers to monetize first‑party signals, not identifiers. He advocated using large language models combined with Retrieval‑Augmented Generation (RAG) to infer reader intent and provide...
Agentic AI 'Cyber Hunting Machines' Coming to Global Real Estate
World Property Markets announced development of three AI-driven platforms aimed at fixing deep structural inefficiencies in the $600 trillion global real estate market. The dual‑platform "Matching Engine" will combine World Property Search, a planetary‑scale listing and buyer‑intent engine, with Global Listings,...
Researchers Discover Major Security Gaps in LLM Guardrails
Unit 42 researchers revealed that safety guardrails in generative AI, termed “AI Judges,” can be bypassed using a novel prompt‑injection technique. Their custom fuzzer, AdvJudge‑Zero, automatically discovers low‑perplexity token sequences that shrink the logit gap between “allow” and “block,” achieving a...
Canva May Have Just Cracked One of the Biggest Issues With AI Images
Canva has launched Magic Layers, a new AI feature that deconstructs any AI‑generated image into editable layers within the Canva editor. Users can upload images from external generators, and the tool identifies backgrounds, text, characters and other elements in under...
SoftBank Takes Aim at Latency with some AI Wizardry
SoftBank unveiled Autonomous Thinking Distributed Core Routing, an AI‑driven system that pairs a CAMARA QoD API with an intelligent agent to dynamically select the optimal network path. The solution toggles between user‑plane‑function routing for efficiency and SRv6 mobile user plane...
HIMSS26: CharmHealth Launches CharmCopilot, an Embedded Multi-Agent AI Assistant for EHR Workflows
CharmHealth has introduced CharmCopilot, an embedded multi‑agent AI assistant built directly into its electronic health record platform. The company is offering the core AI capabilities to existing customers at no additional charge, positioning the tool as a cost‑disruptive solution. Its...
Slashing Agent Token Costs by 98% with RFC 9457-Compliant Error Responses
Cloudflare now returns RFC 9457‑compliant structured Markdown and JSON error payloads to AI agents, replacing the traditional HTML error pages. Agents can request these formats via the Accept header and receive concise, machine‑readable instructions such as retry intervals or escalation guidance....
Netskope Launches AI Security Platform to Monitor and Protect Enterprise AI Systems
Netskope unveiled Netskope One AI Security, extending its Zero Trust platform to monitor and control AI agents, models, and data flows across enterprises. The suite adds an Agentic Broker, AI Guardrails, AI Gateway, and AI Red Teaming to give security...
Broadcom Launches Taurus BCM83640 A 3nm 400G Per Lane Optical DSP
Broadcom unveiled the Taurus BCM83640, a 3nm optical PAM4 DSP delivering 400 Gbps per lane and enabling 1.6 Tbps transceivers. The chip acts as an 8‑to‑4 gearbox PHY, offering best‑in‑class bit‑error‑rate and power efficiency. It paves the way for future 3.2 Tbps modules...
HIMSS26: Imprivata Introduces Agentic Identity Management to Secure and Govern AI Agents
Imprivata unveiled Agentic Identity Management at HIMSS26, a platform that secures and governs AI agents used in clinical documentation, prior authorizations, and triage. The solution treats AI agents as managed identities, issuing short‑lived tokens and enforcing least‑privilege access across both...
From Hype To Outcomes: How VCs Recalibrate Around Agentic AI
The conversation around agentic AI is moving from speculative ambition to concrete outcomes as 2026 approaches. Snowflake's Startup 2026 report, based on interviews with eight AI‑focused VCs, shows investors now value agents that are embedded in specific, data‑rich workflows and...
Hybrid AI Planner Turns Images Into Robot Action Plans
MIT researchers unveiled VLM‑guided formal planning (VLMFP), a hybrid AI system that converts a single image into executable robot action plans. The approach couples a vision‑language simulator (SimVLM) with a generative model (GenVLM) to produce PDDL files for classical planners,...
Google Unveils New AI Cloud Tools to Support Game Development
Google Cloud unveiled a new AI‑driven game development engine at GDC, integrating Gemini 3 Pro, Flash, Gemini Enterprise and Vertex AI. The platform introduces autonomous AI agents that automate repetitive tasks such as testing, code generation, and asset pipelines, while offering a...
The Art of the Handoff: How Agentforce Voice Escalation Works
Agentforce Voice introduces a structured handoff framework that moves beyond the traditional emergency bailouts of voice bots. By default, the system instantly escalates when a user asks for a human, bypassing any pre‑processing. The platform adds a dynamic escalation layer,...
This Digital Picture Frame Wants to Bring People Closer to a Holographic Future
Looking Glass, a Brooklyn‑based holographic‑screen pioneer, unveiled Musubi, an AI‑driven digital picture frame that projects photos and short videos as 3D holograms. The device extracts key subjects from media, displays them across a 170‑degree view, and stores up to 1,000...
Findem AI Hiring Platform Now Available in AWS Marketplace
Findem, an AI‑powered hiring platform, is now listed in AWS Marketplace, allowing enterprises to purchase and deploy the solution directly through their existing AWS accounts. The integration streamlines procurement, consolidates licensing and billing, and lets software spend count toward cloud...
Hustlers Are Cashing in on China’s OpenClaw AI Craze
China’s open‑source AI agent OpenClaw has sparked a grassroots service boom, with engineers like Feng Qingyang turning remote installation support into a full‑time business. By February, Feng’s operation grew to over 100 staff and has processed more than 7,000 orders...
The Download: Pokémon Go to Train World Models, and the US-China Race to Find Aliens
Niantic Spatial, a spin‑out of Pokémon Go creator Niantic, is turning the game’s massive crowdsourced map data into a real‑world model that can ground large language models and give delivery robots centimeter‑level navigation. The firm hopes this will enable autonomous vehicles...
Entrepreneurs Bet on Growth and AI Despite Geopolitical Risks, UBS Survey Finds
The UBS 2026 Global Entrepreneur Report, based on 215 founders across 26 markets generating $34.3 billion in revenue, shows 68% of entrepreneurs remain optimistic despite geopolitical tensions. Over half plan to hire new staff this year and 80% expect headcount to...
AllHere Set Meeting With LAUSD Leaders Months Before Landing $6.2M Chatbot Deal
In January 2023 LAUSD leaders met with AllHere CEO and consultant Debra Kerr, a close associate of Superintendent Alberto Carvalho, months before the district approved a $6.2 million AI‑chatbot contract. The deal later became the focus of FBI raids, an investigation...
U.S. Workers Are Carving a Path to a New American Dream
AI is accelerating as an economic force, but U.S. workforce fragility predates generative models. A Business for Good Foundation survey shows 75% of workers now view a "good job" differently, 80% favor trade training over four‑year degrees, and a majority...