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
Why AI Agents Are Triggering a Rethink of Enterprise Identity
Enterprises are rethinking identity management as AI agents become integral to automation. Traditional token‑based access, which assumes a one‑time authentication, no longer protects against evolving threats. Organizations are moving toward continuous, context‑aware verification that treats humans, workloads, and LLM‑driven agents as first‑class identities. This shift aligns identity with zero‑trust principles, requiring short‑lived credentials, explicit delegation, and real‑time policy enforcement.
Are Your Agents Quietly Draining Your Budget?
Enterprises are rapidly scaling autonomous AI agents, with deployments doubling in 2025 as pilots move into production. Real‑world usage reveals cost spikes up to ten times the prototype budget, often driven by looping or misconfigured agents that can squander tens...
Clinical Utility of the AITIS Model for Test-Free Identification of Sarcopenia in Patients with Stage IV–V Non-Dialysis-Dependent Chronic Kidney Disease
The study validates the Artificial Intelligence to Identify Sarcopenia (AITIS) model as a test‑free screening tool for sarcopenia in 236 patients with stage IV–V non‑dialysis chronic kidney disease. Using only age, sex, height, weight and 20 self‑reported functional items, AITIS achieved...
Development of an Intelligent Food Nutrition Recognition and Nutrient Intake Assessment System in Hospital Settings
Researchers at Peking University People's Hospital created an AI‑driven food nutrition recognition system for therapeutic hospital meals. The platform combines RGB‑D imaging, the Segment Anything Model for segmentation, and a SE_ResNet50_vd classifier, achieving 99.2% accuracy in food type identification. Volume...
Breach Secure Now Gives MSPs an AI Adoption Playbook
Breach Secure Now (BSN) has introduced an AI Risk to Adoption Program aimed at helping managed service providers (MSPs) guide small‑and‑medium business (SMB) customers through safer AI deployment. The partner‑led offering supplies a repeatable model covering risk assessment, shadow‑AI discovery,...
Virtue AI Launches PolicyGuard: Real-Time AI Guardrail Creation with Stackable Compliance and No Engineering Overhead
Virtue AI unveiled PolicyGuard, a real‑time AI guardrail platform that lets enterprises define, edit, and enforce custom AI policies without engineering bottlenecks. The solution translates natural‑language policies, PDFs or JSON into enforceable controls and layers over 30 compliance frameworks such...
Cops, Robbers and Robots: How AI Is Changing Cybercrime
Artificial intelligence is reshaping cybercrime, illustrated by a recent breach where hackers exploited Anthropic's Claude large‑language model to siphon roughly 150 GB of sensitive Mexican government data. The incident highlights a broader trend: AI‑driven "vibe hacking" enables both seasoned and novice...
GM’s AI Story Is Taking Shape as the Carmaker Aims to Distinguish Itself From Rivals Like Tesla
General Motors reported solid first‑quarter earnings, emphasizing the growing role of its autonomous‑driving technology. The automaker highlighted Super Cruise, its advanced driver‑assistance system launched in 2017, as a key differentiator from rivals such as Tesla. Executives said the company sees...
interos.ai Launches iQ to Elevate Supply Chain Risks to the C-Level
Interos.ai unveiled iQ, a second‑generation AI‑driven platform that merges ERP identifiers with its Resilience knowledge graph to deliver predictive supply‑chain risk analytics. The solution quantifies financial exposure, maps tariff impacts, and suggests alternative suppliers, targeting Fortune 1000 enterprises and government agencies....
Weaponized Deepfakes Arrive: AI’s New Existential Threat
Between improvements in generative AI, Grok’s mass generation of nonconsensual sexual images, and a US administration using the technology for propaganda, the long-predicted threat of weaponized deepfakes is here. That's why we named them one of the 10 Things That...
Canonical's Approach to AI Is Refreshingly Thoughtful - Microsoft Should Take Note
Canonical announced that Ubuntu 26.04 will integrate AI through both implicit background enhancements and explicit opt‑in tools, emphasizing local inference and open‑weight models. The company’s strategy avoids rebranding the OS as an AI product, instead offering developers the choice to...
AMD's Lemonade SDK 10.3 Now 10x Smaller By Getting Rid Of Electron
AMD’s open‑source Lemonade SDK 10.3 was released with major upgrades, most notably swapping Electron for the Rust‑based Tauri framework. This change shrinks the pre‑built Windows and macOS binaries from roughly 101‑107 MB to just 7‑9 MB, a ten‑fold reduction. The update also...
Stacklet Joins the Agentic AI Foundation (AAIF) to Advance Open Standards for Autonomous Cloud and AI Infrastructure
Stacklet, creator of Cloud Custodian, has become a Silver Member of the Linux Foundation‑hosted Agentic AI Foundation (AAIF). The move positions Stacklet to help shape open standards for autonomous cloud and AI infrastructure. With its Jun0 platform and newly launched...
AvePoint Doubles Down on Agentic AI and Multicloud Resilience Across the Confidence Platform
AvePoint unveiled major upgrades to its Confidence Platform, emphasizing agentic AI governance and multicloud resiliency. The new Rapid Recovery System shifts disaster recovery focus to restoring a Minimum Viable Company first, while expanding backup coverage to Azure PostgreSQL, Cosmos DB,...
An Oligarch’s Dystopian Scheme to Discredit Journalism with AI
A Peter Thiel‑backed startup, Objection AI, launched this month offering an AI‑driven "jury" that grades journalists with an "Honor Index" and arbitrates media disputes for a $2,000 fee. Founded by Aron D’Souza, the lawyer who led the Thiel‑funded Gawker lawsuit, the service blends...
AI Where It Matters: Why Apple Is Poised to Become the Invisible AI King
Apple is positioning itself as the "invisible AI" leader by embedding Google’s Gemini model into Siri for the upcoming iOS 27 release. Rather than chasing benchmark‑topping models, the company is betting on hyper‑personalized, behind‑the‑scenes AI that deepens its ecosystem moat. This...
Affinity Launches Model Context Protocol to Connect Relationship Intelligence to AI Tools for Private Capital
Affinity unveiled its hosted Model Context Protocol (MCP), linking its relationship‑intelligence CRM directly to AI assistants such as ChatGPT, Claude, Gemini and Copilot. The protocol translates natural‑language queries into secure API calls, enabling private‑capital teams to retrieve, update, and annotate...
Digistore24 Rolls Out MCP Server, Enabling New Layer of AI-Powered Efficiency for Vendors, Affiliates
Digistore24 unveiled its Model Context Protocol (MCP) Server, letting vendors and affiliates link AI agents directly to the sales platform. The new layer enables AI tools such as Claude Desktop, Cursor and Windsurf to manage products, orders, campaigns and reporting...
Wrike Connects Governed AI To Enterprise Workflows, Now Available On OpenAI GPT Store
Wrike announced that its Managed Collaboration Platform (MCP) Server is now listed in the OpenAI GPT Store, enabling enterprise teams to connect ChatGPT directly to Wrike's Work Intelligence Graph. The integration offers OAuth‑secured, user‑scoped access with full audit trails, turning...
AI‑Driven Workforce Demands Immediate Lock‑In Cost Management
Any company that truly believes AI agents will be a significant part of their workforce in the future should be working overtime on the lock-in problem. Both the CTO & CFO should be having sleepless nights about x% of their workforce...
Syntax Launches AI-First Application Managed Services for SAP Solutions
Syntax introduced an AI‑First Application Managed Services (AMS) model to modernize support for SAP environments, covering ECC, S/4HANA, Cloud ERP and line‑of‑business apps. The platform embeds AI into ticket handling, delivering up to 40% reduction in support effort and about...
X Stop Guessing. Start Deciding with Precision.
Legal teams are adopting AI platforms that can review contracts in minutes, but the tools stop at flagging issues without prescribing actions. Recent RFPs and procurement failures reveal a missing “judgment layer” that defines who decides on high‑risk clauses and...
DeepSeek-V4 Is Here. So Is Everybody Else.
DeepSeek released two new models—V4‑Pro (1.6 trillion total, 49 billion active) and V4‑Flash (284 billion total, 13 billion active)—instead of a single V4 offering. Pricing for V4‑Pro is markedly higher than the disruptive rates of V3, pushing effective costs to about $1.74 per million...
Agentic AI Floods Grants, Demands New Funding Models
Agentic AI is rapidly increasing the volume and quality of grant applications, challenging traditional research funding systems and prompting calls for new AI-driven approaches to manage demand and assess quality. aiethics
Data Agents Need Context Graphs. Can Your Data Pipelines Cater?
The article argues that decision traces—approvals, exceptions, and reasoning captured in Slack, email, and workflow tools—are fundamentally events and can be processed with existing behavioral‑event pipelines. By storing these traces in a warehouse‑native context graph, organizations can reuse the same...
Liverpool FC and SAS Will Deliver More Personalized, Real-Time Digital Fan Experiences with AI
Liverpool Football Club has entered a multiyear partnership with analytics firm SAS to embed AI-driven marketing across its digital properties. Using SAS Customer Intelligence 360 and the Viya analytics platform, the club will unify web, mobile and social data to...
AI Investments in Extreme Environment Materials Surpass $262 Million
Industry leaders have pledged over $262 million to accelerate AI‑driven research and production of advanced materials for extreme‑environment applications. Siemens leads the charge with a $259.9 million commitment to expand AI‑enabled R&D hubs across North America, while smaller grants target university‑level AI...
Tesla Begins Probing Owners on FSD’s Navigation Errors with Small but Mighty Change
Tesla’s latest Full Self‑Driving update (v14.3.2, version 2026.2.9.9) replaces the generic “Other” option in the intervention menu with a dedicated “Navigation” label. The change lets owners directly report map‑related disengagements, giving Tesla cleaner data on routing and speed‑limit errors. Navigation...
AI Data Bursts Force Rethink of Data Center Networking at Backblaze
Backblaze is swapping its legacy 100‑gigabit links for 400‑gigabit connections to accommodate the bursty, unpredictable traffic generated by AI workloads on neocloud providers such as CoreWeave and Lambda. The upgrade includes higher‑density Arista switches, allowing terabit‑per‑second internal flows and more...
☕🤖Tutorial: Turn Claude Design Into Your In-House Brand Designer
The post walks readers through converting Anthropic's Claude Design into a private, on‑brand design engine. It outlines a three‑step workflow—custom style‑guide ingestion, prompt engineering, and iterative fine‑tuning—plus a downloadable prompt library. By embedding brand assets and tone, the AI produces...
#AAAI2026 Invited Talk: Yolanda Gil on Improving Workflows with AI
At AAAI 2026, USC professor Yolanda Gil presented how AI can streamline scientific workflows, drawing on ethnographic studies of geoscientists. She highlighted seven recurring problems, such as reliance on familiar methods and fragmented nomenclature, and demonstrated AI solutions using abstraction,...
AI Writes Enterprise Code Fast. Cleaning It up Is Another Story Entirely
AI‑driven code generators can produce Java or Python snippets in minutes, promising rapid application development. However, senior engineering leaders at Appian warn that integrating this code with authentication, data pipelines, and business rules creates layers of complexity that quickly generate...
From the Sistine Chapel to Scaled Intelligence
The article argues that AI success hinges on scaling intelligence through disciplined system design, not merely adding automation. It likens modern enterprises to Michelangelo’s Sistine Chapel, where culture, work structure, and strategy aligned thousands of hands toward a single vision....
Google Exec Argues Safer LLMs, Not Bans, Aid Mental Health
Google executive defends potential of LLMs for mental health. “It can seem sometimes like shutting something down is a way of preventing harm,” Megan Jones Bell told STAT. “We believe that making our product experience safer and more helpful and...
A Solo Founder with Claude Beats Ten‑person Teams Lacking It
A solo founder with Claude set up correctly is more dangerous than a 10-person team without it. Most people I talk to are using Claude for one-off questions. Write me an email. Fix this paragraph. Give me three taglines. The output...
Livestream Today: Building Supercompanies
Scott Galloway announces a live‑stream at 1:30 p.m. ET with Greg Shove, CEO of Section, to discuss "Building Supercompanies." A Supercompany is defined as an organization that converts AI adoption into business value faster than competitors, attracting superior capital, talent, and...
Bipartisan Bill Targets Deepfakes, Tightens AI Regulations
Lawmakers are moving to address AI risks more directly. A new bill would impose stricter penalties for distributing deepfakes and establish clearer standards for AI systems, with bipartisan backing. It reflects a broader shift. Regulation is starting to catch up with the...
Apptronik Poaches Top Robot Talent Amid Growing War
The humanoid robot talent wars are heating up ... >> Apptronik Teases New Robot, Stacks C-Suite With Waymo, Boston Dynamics, Amazon Vets https://t.co/uPY9OnVW7O
Our Deal Management Agent: Bringing Scaled AI-Adoption to PMPs, PGs, and Auction Packages
PubMatic’s Deal Management agent, launched last August, automates the configuration of private marketplace (PMP) and programmatic guaranteed (PG) deals using natural‑language prompts. The AI tool has already powered over 1,000 deals, cutting setup time by roughly 87% and handling millions...
Microsoft Backs Agent-to-Agent Protocol in New Framework
Glad to see Microsoft betting on the Agent-to-Agent Protocol for their v1 Agent Framework (Python/C#) ... https://t.co/3R39hl934s
Ubuntu Prioritizes Hardware Support and Local‑First AI
I started to talk with teams building the leading OSes about how AI changes (or does not change) their roadmap and what their OS looks like. Ubuntu is one that feels very grounded: massive focus on hardware support (GPUs, NPUs, DPUs)...
GM Hits 1 Billion Hands-Free Miles, Will Drop Google Gemini Into 4 Million Cars
General Motors announced that its Super Cruise driver‑assistance system has accumulated 1 billion hands‑free miles across a fleet of roughly 750,000 North American vehicles. The automaker expects paid subscriptions to exceed 850,000 by year‑end, with about 40% of users renewing after...
AI Demand Surpasses Supply; WSJ Spreads Fear
Demand outstrips supply everywhere in AI. That WSJ piece was a giant FUD burger. 🍔
AI Prompts Yield Inconsistent Analysis Results—Reproducibility Matters
I ran the same AI prompt on the same dataset twice. Got different figures in the report both times. Here's why that matters if you do data analysis with AI agents. https://t.co/YL4muRdkL2
The Manus Doctrine
China’s National Development and Reform Commission ordered Meta to unwind its $2 billion purchase of AI startup Manus on April 27, 2026. Manus, founded in Beijing in 2022, had moved to Singapore after a $75 million funding round in mid‑2025 and was integrated...
AI Reads Retinal Images to Screen Multiple Diseases
What superhuman vision can detect from the retinal photo, which human eyes cannot, is stunning. A new foundation AI model screening for diabetes hypertension, hyperlipidemia, gout, osteoporosis, and thyroid disease @NatureMedicine https://t.co/GhKvUqz4Vy https://t.co/iKcXCbLceu
LinkedIn Ready for AI Citations, Video Booms in B2B SaaS
I'll say it again: @LinkedIn is ripe for improving AI citation rate for B2B SaaS + video is TAKING OFF there. https://t.co/E7xNdgqwyj
Free Webinar on How To Use AI To Sell More Tickets
Bauer Entertainment Marketing is hosting a free webinar titled “AI Built For The Business Of Selling Tickets” on April 29 at 10:30 AM PT/1:30 PM ET. The session will demonstrate how artificial intelligence can be applied to live‑event marketing tasks such as localizing...
New Research Restores Original Sensor Data From AI-Edited Photos
Restoring What Your Camera Captured Before AI Changed It 📸 💬 How can you protect the sanctity of a raw photograph from AI interference when it’s already been automatically run through AI inside the camera? New research seeks to restore...
Snapchat Brings AI-Powered Conversational Advertising to Its App
Snapchat is rolling out AI Sponsored Snaps, interactive ads placed in the app’s Chat tab that let users converse with brand‑run AI agents. The format builds on existing Sponsored Snaps, which already deliver 22% higher conversion rates and roughly 20% lower cost‑per‑action....