Today's AI Pulse
Anthropic shuts down Claude Fable 5 and Mythos 5 after US export‑control warning
Anthropic disabled access to its Claude Fable 5 and Claude Mythos 5 models after the Trump administration, via the Commerce Department, warned that foreign nationals could exploit the systems. The move followed alerts from Amazon and other firms about a potential jailbreak that could aid cyber‑security attacks. Anthropic declined the administration’s request for a voluntary shutdown but complied with the order.
Also developing:
By the numbers: AMD acquires AI memory‑optimization pioneer Mext
AI Alone Won’t Change Your Business. The System Running It Will.
Microsoft announced a comprehensive enterprise AI agent platform that unifies building, contextualizing, running, governing, and continuously improving AI agents. The system leverages Azure, GitHub, Microsoft IQ, Foundry, Agent 365, and the broader Microsoft security stack to deliver a single, integrated environment. By embedding governance, identity, and compliance from the start, the platform aims to make AI agents trustworthy and scalable across functions like finance, HR, and software delivery. Continuous feedback loops and Frontier Tuning enable agents to evolve safely, turning AI from isolated demos into a core operating system for business work.
Securing AI Agents Before They Go Rogue Is Next to Impossible
At the Gartner Security & Risk Management Summit, Dennis Xu warned that fully securing high‑autonomy AI agents is practically impossible. He highlighted the PocketOS incident where an AI coding agent deleted a production database and its backups in nine seconds...
Microsoft’s Vasu Jakkal On Why AI Agents Need Human-Level Security Controls
Microsoft announced an expansion of its Agent 365 platform to include local AI agents, moving the service from cloud‑only to hybrid environments. The preview adds a registry that discovers and classifies more than 20 types of on‑premise agents, giving security teams...
New Microsoft Tool Lets Devs Spin up AI Behavior Tests Using Text Descriptions
Microsoft unveiled ASSERT, an open‑source framework that converts natural‑language policy descriptions into automated, scored tests for AI models. The tool generates test scenarios, runs them against target systems, and records decision paths to pinpoint failures. ASSERT supports pre‑deployment validation, post‑deployment...
The Agentic Reckoning: Enterprise AI Organizations Have a Runtime Problem, Not a Model Problem — and Most Are Building the...
VentureBeat’s Q1 2026 Pulse survey of 132 senior technology leaders finds enterprise AI failures stem from runtime infrastructure, not model quality. Stateless Python‑based agents lose context, incur token overruns, and propagate hallucinations, exposing a “spine” bottleneck in state management and...
Machine Learning Model Accurately Predicts Long-Term Risk of Type 2 Diabetes
A new electronic‑health‑record based machine‑learning model predicts a patient’s risk of developing type 2 diabetes up to ten years in advance. The retrospective study analyzed 3,365,464 adults from Kaiser Permanente Northern California, achieving an area‑under‑the‑curve of 0.886 in training and 0.883...
Microsoft's First Reasoning Model Is One of 7 AIs Just Released at Build - What We Know so Far
Microsoft announced seven new AI models at its Build conference, highlighted by MAI‑Thinking‑1, the company’s first 35‑billion‑parameter reasoning engine trained on enterprise‑grade, commercially licensed data. Independent blind tests showed the model outperformed Anthropic’s Sonnet 4.61, while MAI‑Code‑1 will power Copilot and...
Gusto Introduces an ‘AI Teammate’ for Small Businesses
Gusto unveiled "Gusto Cofounder," an AI‑powered teammate designed for small‑business payroll and operations. The tool embeds directly in Gusto’s platform, pulling in existing team, payroll schedule, and calendar data so it can act without any initial setup. Users can trigger...
New Bill Aims to Regulate Military Uses of AI
Senator Kirsten Gillibrand introduced the Secure and Accountable Military AI Act, aiming to restrict the Pentagon’s use of artificial intelligence in high‑risk domains such as nuclear weapons, lethal targeting, domestic surveillance, and fully‑autonomous systems. The bill would require senior defense...
"Agents Are only as Good as the Context We Give Them": Microsoft IQ Connects AI Agents to Your Workspace Data...
Microsoft unveiled Microsoft IQ at the Build conference, a new context layer that grounds AI agents with enterprise data and web knowledge. The platform comprises Work IQ, Fabric IQ, Foundry IQ and Web IQ, each adding semantic understanding, data integration, or faster search. Work IQ APIs...
Hermes Desktop Takes AI Agents Off the Browser
Most AI agents still live inside a browser. Hermes Desktop changes that. First revealed during Jensen Huang's GTC keynote, Hermes is now running natively on your machine, bringing faster performance, deeper system access, and a more seamless AI experience. This is what the...
Microsoft Offers Devs a Better Way to Control AI Agent Behavior
Microsoft unveiled the Agent Control Specification (ACS), an open‑source framework that lets developers codify granular policies for AI agents. The spec defines what agents may do, must avoid, when human approval is needed, and what audit logs to capture, applying...
Google Phone App Rolling Out Android Fake Call Detection that Uses RCS
Google’s Phone app will roll out a fake‑call detection feature that leverages RCS to verify callers in real time. The system sends a silent encrypted signal between Android devices; if the signal is missing, users receive a warning. The feature...
Neurosymbolic AI Complements Deep Learning, Proving 2022 Prediction
People who say this kind of thing are completely lost about what I actually said about deep learning, and I would strongly encourage them to read “Deep learning is hitting a wall” (2022). What I said there (and in 2018...
Trump Signs Order Seeking Govt Access to New AI Releases
President Trump signed an executive order creating a voluntary framework that requires AI developers to share their most advanced models with the U.S. government before public release. Companies such as OpenAI, Google and Anthropic must provide a 30‑day window for...
Trump Signs Executive Order to Promote Advanced AI Innovation and Cybersecurity
President Donald J. Trump signed an executive order on June 2, 2026 to accelerate cyber defenses and create a voluntary framework for the secure deployment of advanced, "frontier" AI models. The order mandates 30‑day and 60‑day deadlines for agencies to...
Steven Sinofsky on AI PCs, NVIDIA, and the Future of Computing
Steven Sinofsky, former Windows division president and creator of the Surface line, discusses the recent Computex announcements, notably NVIDIA's RTX Spark (N1X) ARM‑CPU/GPU chip aimed at AI‑native PCs. He explains how AI workloads are shifting compute from CPUs to GPUs...
Bloomberg Tech: Anthropic Files For IPO (Podcast)
Alphabet announced plans to raise roughly $80 billion in equity to fund a massive expansion of its AI‑infrastructure platform. At the same time, Anthropic filed a confidential registration statement, positioning itself ahead of OpenAI in the race to go public. SpaceX...
Nvidia Wants to Supercharge Your Laptop
Nvidia announced plans to bring its AI‑centric hardware and software stack to consumer laptops, extending the same performance thrust that has driven its data‑centre dominance. CEO Jensen Huang highlighted a forthcoming line of GPUs and developer tools designed to run...
SymphonyAI Launches CINDE Retail Media Intelligence
SymphonyAI unveiled CINDE Retail Media Intelligence (RMI), an AI‑driven platform that fuses merchandising, media and shopper intelligence into a single closed‑loop system. The solution plugs into existing retail‑media stacks and partners with third‑party measurement firms to deliver audience segmentation, causal...
Microsoft Releases Web IQ, Powered by Bing but Designed for How AI-Agents Search
Microsoft unveiled Web IQ, a new suite of AI‑native grounding APIs that connect large language models and autonomous agents to up‑to‑date web content, including pages, news, images and videos. Powered by Bing’s index, the service is engineered for the “agentic...
Microsoft Build 2026: Building Agentic Apps with Microsoft Fabric and Microsoft Databases
Microsoft Build 2026 unveiled new tools to accelerate AI‑driven app development on its unified data platform. Rayfin, an open‑source SDK and CLI, lets developers and coding agents generate production‑grade backends on Microsoft Fabric with built‑in security and scaling. Azure HorizonDB,...
Build Meaning Before Machines: Why Semantics, Ontologies, And Knowledge Graphs Matter For Agentic AI
Enterprises are discovering that agentic AI systems stumble when data lacks explicit meaning. Without semantics, ontologies, or knowledge graphs, autonomous agents misinterpret joins, metrics, and context, leading to flawed actions. Recent Forrester reports argue that a semantic layer is the...
How AI Is Reshaping Cybersecurity in Utility Operations
Artificial intelligence is rapidly reshaping utility cybersecurity, sparking an arms race where both attackers and defenders deploy AI to discover vulnerabilities and launch sophisticated attacks. Utilities are countering with AI‑driven anomaly detection, generative tools, and agentic systems that can autonomously...
Call My Agent: Hollywood’s Data Wants a Deal
The Ankler outlines how Hollywood’s vast content libraries, long scraped by AI developers, are now being packaged by emerging data brokers who sell ethically sourced, rights‑cleared training datasets. Recent lawsuits—Disney and NBCUniversal against Midjourney, and OpenAI’s Sora controversy—highlight studios’ concerns...
New Azure Cobalt 200 VMs Deliver 50% Performance Improvement, Fully Optimized for Modern Agentic AI Workloads
At Microsoft Build 2026, Azure announced an early‑access preview of Cobalt 200 Arm‑based virtual machines, the second‑generation of its custom silicon. Cobalt 200 delivers up to 50 % higher CPU performance, 20 % more remote‑storage IOPS and 15 % greater network bandwidth versus Cobalt 100, and...
Microsoft Debuts Surface RTX Spark Dev Box to Run Large AI Models without Cloud Costs
Microsoft unveiled the Surface RTX Spark Dev Box, a compact desktop powered by Nvidia's Blackwell‑based RTX Spark processor and 128 GB of unified memory. The machine delivers roughly one petaflop of AI compute, enabling developers to run models over 120 billion parameters...
Microsoft Launches MXC, an OS-Level Sandbox for AI Agents, with OpenAI and Nvidia Already on Board
Microsoft unveiled Microsoft Execution Containers (MXC) at Build 2026, an OS‑level sandbox that lets developers and IT admins define precise policies for AI agents, enforced by the Windows kernel. MXC ships as an SDK and policy model built into Windows...
Anthropic’s Office Launched an AI-Run Vending Machine. It Evolved Into AI-Run Stores and Cafes Within a Year
Andon Labs, in partnership with Anthropic, upgraded an AI‑run vending machine into fully autonomous retail stores and cafés within a year. The multi‑agent system, led by a mechanical CEO, handles hiring, procurement, logistics and even passes stringent Swedish labor inspections...
Nvidia Pours $1 Billion Into Nokia to Fast‑track AI‑RAN and 6G Roadmap
Nvidia has committed a $1 billion strategic investment in Nokia, earmarked for AI‑enabled radio access network (AI‑RAN) development and the transition toward AI‑native 6G. The deal follows a surge in Nokia’s AI‑related revenue, which grew 49% in Q1, and has sparked...
Meta’s AI‑Powered Advantage+ Catalog Ads Cut CPA by Up to 38%, Prompting Creative Overhaul
Meta’s new AI framework, anchored by Advantage+ Catalog Ads, is delivering 30‑38% lower cost‑per‑acquisition for top DTC brands. The shift compels advertisers to replace narrow audience stacks with broad targeting and diversified creative assets.
TSMC Says Energy Efficiency, Not Raw Power, Is AI Chip Priority
TSMC senior vice‑president Kevin Zhang told reporters in Amsterdam that AI chip buyers are now prioritising energy efficiency over sheer horsepower. The foundry aims to slash power consumption by as much as 30% between its N2 and A14 nodes while...
Naver Pushes AI Shopping Agent to Initiate Purchases, Redefining Korean E‑Commerce
South Korea’s Naver upgraded its AI Shopping Agent on June 1 to proactively push product suggestions on the Naver Shopping app, moving from a reactive search model to a push‑commerce approach. The change follows three months of beta testing and aims...
Cognition AI Raises $1 Billion as Devin Writes 90% of Its Code
Cognition AI closed a Series D round of more than $1 billion, pushing its post‑money valuation to $26 billion. The funding underwrites the rollout of Devin, an autonomous coding agent that now produces roughly 90% of the company’s own software, highlighting a...
MemryX Appoints Joe Faris as VP of Sales and Marketing to Accelerate AI Infrastructure Rollout
MemryX Inc. announced the appointment of Joe Faris as Vice President of Sales and Marketing, bolstering its leadership as the company expands AI inference hardware from edge deployments into data‑center opportunities. The move pairs Faris with new CEO Ross Jatou,...
How to Secure AI Agents: A Practical Overview for Development Teams
Docker’s practical guide warns that 45% of organizations struggle to secure AI agents, which are moving to production faster than security practices can keep up. Because agents autonomously select tools, chain actions, and retain memory, traditional static controls fall short....
SAP Says Enterprise AI Wins Require Context‑Rich Agents, Not Bigger LLMs
SAP announced its Business AI Platform at Sapphire 2026, emphasizing that delivering business context to AI agents, not owning the largest language model, will determine enterprise AI success. The strategy leans on open‑source frameworks and partner LLMs while leveraging SAP’s...
'People Talk About AI Reducing Jobs, Complete Nonsense': Nvidia's Jensen Huang Criticises Economic Doomerism on GTC Stage
At Nvidia’s GTC in Taipei, CEO Jensen Huang declared that "useful AI" is now delivering real ROI and argued that fears of AI‑driven job losses are unfounded. He cited the sector’s 30 million software developers, who generate roughly $3 trillion in salaries...
Mecka AI Secures $60M to Train Robots with Human Sensor Data
Mecka AI closed a $60 million financing round—$25 million Series A in November and a $35 million follow‑on—led by Framework Ventures. The New York‑based startup will use the capital to scale its human‑generated sensor data platform that trains robots for commercial use.
STMicroelectronics Raises FY26 Data‑Center Revenue Forecast to $1 B, Eyes Doubling FY27
STMicroelectronics announced it now expects FY26 data‑center revenue of about $1 billion, up from just over $500 million, and projects FY27 could double that figure. The upgrade follows robust AI‑driven demand and a capacity ramp‑up, sending the stock up 2.3% in after‑hours...
Recommendation Algorithms Might Be Making Your Entertainment Boring, New Research Suggests
A new study in the Journal of Cultural Economics finds that highly accurate content recommendation algorithms can make entertainment feel boring over time. The research, led by University of Toronto assistant professor Samsun Knight, uses a dynamic mathematical model to...
Foundry IQ: Build Smarter Agents Faster with Unified Knowledge and Serverless Retrieval
Microsoft unveiled a suite of Foundry IQ enhancements aimed at accelerating enterprise AI agent deployment. The preview of Foundry IQ Serverless introduces a pay‑per‑use model that scales to zero when idle, while new knowledge sources—Work IQ, Fabric IQ, Azure SQL,...
Cisco Delivers Sovereign, AI‑ready Infrastructure for Secure Control
The new imperative via @ChuckRobbins "sovereign infrastructure, where you want us to meet your evolving needs around sovereignty of data, sovereignty of AI, sovereignty of cloud services, and this means that we have to change how we operate to...
New Model Illuminates AI Infrastructure Economy Landscape
This model we developed turned out to be one of the more interesting lenses to analyze those in the AI infra economy. Hypers, neos, and power REITs. And soon for subs there will be a dashboard/simulator to interact with...
ASOS’s Melissa Lim Breaks Down New AI Stylist App Experience
ASOS has launched an AI‑powered Stylist app inside ChatGPT for UK and US shoppers, letting users browse, receive visual styling advice and click through to purchase on ASOS.com. The feature embeds high‑resolution images, video and livestream content directly into the...
AI Streamlines Research: Assumptions to Verified Data Models
Text-to-modeling This is probably the most common use case for AI in my research process right now. I describe my assumptions, ask for a few years of historical data, and make sure I can verify every cell with a single click.
Microsoft Unveils Seven New AI Models Today
Microsoft AI CEO @mustafasuleyman on stage talking about 7 new Microsoft AI models being launched today https://t.co/vZlGWsXAW5
Microsoft Unveils MAI-Thinking-1, Advanced Reasoning AI for Math and Coding
Microsoft is announcing MAI-Thinking-1 today, its first advanced reasoning AI model. It's built for math and coding primarily, and is part of 7 new AI models Microsoft is launching today. Details 👇 https://t.co/Sz5CSCbYQD
Trump Orders Early Government Access to Corporate AI Models
JUST IN: Trump signs executive order asking companies to provide early government access to AI models
Open-Weight Models Claim 69% of OpenRouter Traffic
Open-weight models have overtaken closed models on OpenRouter. 69.1% of token volume now goes to open-weight models. 30.9% to closed. Competition is a discovery procedure — and developers are discovering the value of open models. 🧵 https://t.co/9ZOLA6llvZ