Today's AI Pulse
US export controls force Anthropic to pull Fable 5 and Mythos 5 after jailbreak discovery
Amazon’s security team uncovered a jailbreak in Anthropic’s newly released Fable 5 model that let users bypass safety safeguards. CEO Andy Jassy raised the issue in a call with White House officials, prompting Treasury Secretary Scott Bessent to alert the Commerce Department, which imposed export controls that forced Anthropic to withdraw Fable 5 and Mythos 5 within days.
Also developing:
AI-Generated CSAM: Staying Ahead Of The Threat
The Internet Watch Foundation reported a record 312,000 confirmed CSAM cases, with AI‑generated, photo‑realistic abuse videos soaring by more than 26,000% in a single year. Law‑enforcement investigators now confront deepfake content that cannot be detected through traditional hash‑matching, forcing a shift in forensic methodology. Cellebrite highlights its Pathfinder platform, which leverages facial similarity and image analytics to locate AI‑fabricated material without relying on hashes. The article also notes divergent state legal frameworks and the need for coordinated policy responses.
Cassava Plans AI Factory in Joburg, as Cape Town Goes Live
Cassava Technologies is set to launch its first AI factory in Cape Town within weeks, leveraging Nvidia’s AI platform, and is already planning a second 20 MW AI factory in Johannesburg. The facilities will deliver high‑performance computing, GPU‑as‑a‑service and AI‑as‑a‑service to...
AI Godfather Yann LeCun's Advice on College, Work and Breaking Through AI Hype
Yann LeCun, the Turing‑award‑winning AI pioneer, cautions that AI doom narratives are inflating fear among teens and distorting public policy. He argues CEOs of AI labs exaggerate risks to sell products, and that the claim AI will wipe out 20% of...
As Formula One Evolves, AI Becomes Part of the Race
Artificial intelligence is moving from branding to core operations in Formula One, with eight new AI partnerships signed in the past six months. Williams has teamed with Anthropic’s Claude model to enhance race‑strategy, while Red Bull works with Oracle and McLaren has...
Unified Identity Layer Key to Securing Agentic AI
Silverfort will headline the ITWeb Security Summit 2026 in Johannesburg, urging firms to replace fragmented point solutions with a unified identity security layer. The company highlights two pressures: outdated legacy infrastructure that was never fully secured, and the emergence of...
Spring Cleaning for Your AI Tools
AI chatbots like Anthropic's Claude now retain a rolling memory of prior conversations, updating it every 24 hours to provide contextual continuity. Users can disable this memory, but doing so strips the model of useful context. Similar persistent‑memory features exist...
AI Can't Fix Healthcare's Deep Structural Problems
AI is being positioned as a fix for healthcare. It won’t be that simple. Automation may reduce costs and speed up diagnoses, but it doesn’t solve deeper issues like staffing, funding and system complexity. The risk is clear. Technology can improve parts...
I Let AI Look at My Breasts—And I’m Glad I Did
Wall Street Journal columnist Joanna Stern let an artificial‑intelligence system read her mammograms after years of early screening due to dense breast tissue. The AI flagged subtle patterns that radiologists might miss, offering a personal case study of how machine‑learning...
US Chip Ban Fuels China’s Homegrown AI GPU Surge
The US export ban on Nvidia chips to China didn't slow China's AI one bit; it is building it. Jensen Huang just confirmed what anyone watching Asia already knew: Nvidia's China market share is now 0%. Gone. How exactly the US...
AI Chipmakers in Korea, Taiwan Drive Asian Stocks to Record
Asian equities surged as AI‑focused chipmakers in South Korea and Taiwan propelled the MSCI Asia Pacific Index to a 2% gain, its highest level since April. TSMC, Samsung Electronics and SK Hynix all posted record‑high shares, lifting regional tech benchmarks...
Artificial Stupidity in the Persian Gulf
The United States backed a $3.3 GW AI compute corridor in the Persian Gulf, promising cheap $0.10/kWh power, low‑latency connectivity and up to 70,000 Nvidia chips for American firms. The plan was intertwined with a $2 billion Trump‑linked stablecoin deal and a...
AI Tools Aren't Training; Real Competence Needs Practice
Professional athletes train most of the time to perform when it matters. Corporate teams perform all the time and get one AI workshop called “transformation.” What worries me is how often we confuse access with ability. Giving people AI tools is not the...
The AI Knowledge Gap We Can’t Afford to Ignore
Healthcare is rapidly integrating artificial intelligence, with two‑thirds of physicians using AI tools in 2024—a 78% surge from previous years. While AI can streamline chart review and surface clinical trends, experts warn that overreliance creates automation bias, magnifying documentation flaws...
Small Language Models: Rethinking Enterprise AI Architecture
Enterprises are reshaping AI stacks by routing routine queries to 1‑7 billion‑parameter small language models (SLMs) while reserving trillion‑parameter large language models (LLMs) for complex reasoning. This division of labor can slash cloud inference costs by up to 90 % and deliver...
Agentic AI, Fraud and the Fight for Customer Loyalty
Agentic AI is reshaping e‑commerce by automating the entire buyer journey, from product discovery to checkout. At the same time, fraud‑as‑a‑service is proliferating synthetic identities, deepfakes, and document forgeries, overwhelming traditional safeguards. Merchants must shift to proactive, data‑driven fraud prevention...
Agentic Browsers Rewrite the Rules of Enterprise Security
Enterprise browsers are evolving from passive tools to autonomous agents, driven by rapid AI adoption. Deloitte reports 74% of organizations will deploy agentic AI within two years, while 84% of knowledge workers are eager to use it. These agentic browsers...
What Is Your AI Drug Repurposing Strategy Missing?
The article argues that AI‑driven drug repurposing for oncology often fails because models are fed fragmented, noisy data despite abundant datasets. It stresses that more data alone won’t improve outcomes; instead, high‑quality, curated, structured data—such as knowledge graphs linking genes,...
Global Payroll Is a Strategic Problem. AI Is Finally Solving It.
Global payroll has long been treated as a fragmented, operational chore, relying on a patchwork of local vendors and costly reconciliation. Recent AI advances shift payroll from assistance to autonomous execution, delivering "zero‑click" processing that cuts workforce‑management expenses by roughly...
The AI Gap in Retail Operations Is Real. Here’s How to Start Closing It.
Retail executives overwhelmingly acknowledge that manual back‑office tasks are eroding sales and customer experience, with 72% citing operational drag. A ServiceNow‑Informa TechTarget survey reveals integration hurdles (54%) and unclear KPIs (67%) as the chief obstacles to AI adoption. Yet modest,...
#358 How AI Agents Will Work While You Sleep | Ruslan Salakhutdinov, Professor at Carnegie Mellon
In this episode, DataCamp talks with Carnegie Mellon professor Ruslan Salakhutdinov about the rapid evolution of AI agents, especially their growing capabilities in coding, computer‑use tasks, and long‑horizon problem solving. Salakhutdinov explains how agents are moving from minute‑scale actions to...
Why Your First AI Hire in Finance Should Focus on Operations
The article argues that a finance team’s first AI hire should target operational tasks such as reconciliations, invoicing and cross‑system handoffs rather than analytics. Recent improvements in large‑language‑model precision, guardrails and audit trails now make AI reliable enough for repetitive,...
How OpenClaw’s Agent Skills Become an Attack Surface
OpenClaw’s AI agent offers deep integration with a user’s local machine, granting access to files, browsers, and long‑term memory, but it stores configuration and credentials in plain‑text files. This design lets attackers who compromise the host quickly exfiltrate API keys,...
Data, AI, and the Quest for Edge in Hedge Funds
S&P Global Market Intelligence is positioning itself as the backbone of hedge‑fund data infrastructure, emphasizing accuracy, structure, and trust over sheer volume. Michael Patton outlines four competitive pillars—unique data, speed, interpretation, and behavioral discipline—that drive investment edge. The firm tackles...
Contingency Frameworks for Future U.S.-China Cooperation on AI Assurance and Security
RAND’s latest expert‑insights paper proposes a contingency framework for future U.S.–China cooperation on AI assurance and security. While current geopolitical rivalry makes bilateral AI projects unlikely, the analysis outlines five strategic lines of effort that could enable stable, predictable collaboration...
Self‑hosted Immich Replaces Paid Cloud Photo Services
Google Photos charges $3/month after 15GB. iCloud charges $1/month and scans your images. Amazon Photos bundles you into Prime. Someone open-sourced a Google Photos clone that runs on your own hardware for $0. And no company can delete your memories. It's called Immich. Here's what...
AI Moves to Frontline, Shaping Bank Decisions
AI is moving from back office to front line in banking. Bank of America is deploying AI agents to support advisers in real time, shaping client interactions and recommendations. The shift is structural. AI is no longer assisting processes, it is starting...
Open-Weight AI Models Require Proportional Evaluation Approaches
Open‑weight AI models (OWMs) are emerging with publicly accessible parameters, creating risk profiles that traditional closed‑weight evaluation methods overlook. Researchers Paskov, Rodriguez, Dev, and Casper propose a proportional evaluation (PE) framework comprising four criteria to address these gaps. Their systematic...
XAI Unveils Affordable Grok 4.3 with Rapid Voice Cloning
xAI launches Grok 4.3 at an aggressively low price and a new, fast, powerful voice cloning suite 📣 👉 https://t.co/fT7JvuY5Ws https://t.co/0CuJJNMyT8
Compute, Not Intelligence, Becomes AI’s New Bottleneck
The real bottleneck in AI is no longer intelligence. It’s compute. Rate limits on tools are early signals that chips, power and data centers are under pressure as demand explodes. The implication is strategic. Access to compute may decide who leads and...
MAS Partners Banking Industry to Tap AI, Machine Learning to Combat Financial Crime
The Monetary Authority of Singapore announced a proof‑of‑value project to use artificial intelligence and machine learning for pre‑emptive scam detection. The initiative brings together five local banks, the Government Technology Agency and the Singapore Police Force, creating a secure data‑sharing...
Blackstone, H&F, Goldman Sachs Eye $1.5B Anthropic AI JV
Blackstone, Hellman & Friedman, and @GoldmanSachs near $1.5bn AI joint venture with @AnthropicAI Read more here: https://t.co/08fJYiyLMs https://t.co/cicFdQqZB8
Silicon Valley Made AI Powerful. Tokyo Wants to Make It Work
Tokyo is positioning itself as the world’s most startup‑friendly city, leveraging its stable infrastructure, government backing, and a burgeoning AI ecosystem. Governor Yuriko Koike’s agenda targets a ten‑fold rise in startups and unicorns, and the city has already climbed to...
Enhans Renames CommerceOS to AgentOS, Expanding Its AI Agents Beyond Commerce to the Enterprise
Enhans announced that its CommerceOS platform will be renamed AgentOS effective May 1, signaling a shift from a commerce‑focused tool to a broader enterprise AI operating system. AgentOS promises to stitch together data, ontologies, agents, workflows and web interfaces into a...
Your AI CV Just Got You Rejected
A federal court in California authorized a collective‑action notice allowing anyone over 40 who applied through Workday’s AI hiring platform since September 2020 to join a class‑action lawsuit. The case, Mobley v. Workday, alleges that the AI system systematically down‑weights older,...
Zensai Introduces Human Success Agent for Microsoft Agent 365, Redefining AI-Driven Learning and Human Success in the Flow of Work
Zensai announced the general availability of its Human Success Agent, a partner AI agent built on Microsoft Agent 365. The solution embeds real‑time workforce performance, engagement and skill data directly into Microsoft 365 apps, leveraging Microsoft Entra for secure governance....
How TeamViewer ONE Transforms IT Operations From Firefighting to Autopilot
TeamViewer ONE leverages agentic AI to shift IT support from reactive ticket‑based firefighting to proactive, autonomous remediation. By continuously monitoring endpoints, the platform detects early signals of digital friction and resolves issues before users notice them. The solution combines AI,...
€4.8m EU-Funded AI Project to Tackle Child Diarrhoeal Disease in Africa
An EU‑funded CARE‑AFRICA project will develop an AI‑driven tablet tool to identify the pathogen behind diarrhoeal disease in children under five across sub‑Saharan Africa. The €4.8 million (about $5.2 million) grant brings together six partners from Europe and Africa to train models...
AmEx CEO Stephen Squeri - Why Agentic Commerce Is More Fraught with Risk than Traditional E-Commerce
American Express CEO Stephen Squeri said the firm’s AI rollout has already delivered a 30% productivity lift for its programmers, accelerating product development across its global network. He unveiled the ACE (Agentic Commerce Experiences) Developer Kit, enabling merchants to embed...
OpenAI To Extend Cyber Program to Government Agencies
OpenAI released a roadmap titled "Cybersecurity in the Intelligence Age" outlining plans to broaden its Trusted Access for Cyber (TAC) program to federal, state and local government agencies, as well as key industry players. The expansion coincides with the launch...
AI Empowers Lean Firms to Outrun Chaotic Giants
It’s not a tech issue. Most businesses are chaotic black boxes that just don't know what they want... and you can't optimize what you don't understand. Lean, self-aware companies will use AI to CRUSH large, unwieldy ones. great read👇
OpEd: Artificial Intelligence Redefines Workspace
Los Angeles executives are shifting AI from pilot projects to enterprise‑wide deployment, with 92% naming it a top investment priority for 2026. The surge in AI use is reshaping task assignment, review, and governance, prompting leaders to balance speed with...
Scaffolding, Not Model Size, Is the Real AI Moat
everyone is talking about the `Agent Harness` right now. this is a great breakdown of why: → Frontier models are converging and getting way cheaper. → the real moat is the scaffolding around the LLM (context, tools, sandboxing etc.) 👇
Can Investors Trust AI Sales Figures? Asks Wall Street Journal Opinion Piece
A Wall Street Journal opinion piece warns that AI firms are inflating sales by paying partners to adopt their technology. It highlights OpenAI’s $1.5 billion joint venture with private‑equity firms, Anthropic’s $200 million contribution to a similar deal, and Google’s $750 million subsidy...
The EU AI Act Newsletter #101: Trilogue Breakdown
EU legislators failed to secure a deal to postpone key provisions of the EU AI Act until December 2027, leaving the August 2024 start date for high‑risk AI applications unchanged. The deadlock centers on whether machinery and medical‑device sectors should...
AI-Driven Risk Intelligence: How FIs Are Predicting Systemic Shocks
Financial institutions are adopting AI‑driven risk intelligence to monitor systemic signals in real time, replacing slow, periodic stress‑test cycles with continuous analytics. The European Systemic Risk Board’s 2025‑2026 report highlights AI’s dual nature: it can improve decision‑making and early‑warning capabilities,...
My AI Learning Journey – Part 10 – AI Assisted Coding – VSCodium and Continue
The author integrated the open‑source VSCodium editor with the Continue extension, OpenRouter, and the GPT‑5.3‑codex model to create a fully private AI‑assisted coding workflow. By prompting the LLM, a 600‑line Python utility for monitoring a Fritzbox router was generated in...
Walmart Could Be Prepping a Budget Smart Speaker with Gemini Baked In
Walmart is poised to launch a budget Onn smart speaker that embeds Google’s Gemini generative AI, marking the first third‑party device to run the technology. The CSA filing shows the speaker will support Google Cast for Audio, Wi‑Fi, Bluetooth, a...
Vocus and Fortinet Fighting Shadow AI with Secure Shield Launch
Australian telecom Vocus and cybersecurity firm Fortinet have launched Vocus Secure Shield, a managed SASE platform designed to expose and control shadow AI usage within enterprises. The service leverages Fortinet’s AI‑driven security and deep application awareness to detect AI tools,...
Infios Launches New AI Capabilities Focused on Supply Chain Workflows
Infios unveiled a suite of AI agents that embed directly into supply‑chain workflows, spanning order management, warehousing and transportation. The agents automate tasks such as driver check‑calls, document parsing and inventory research, while dynamically rerouting shipments when disruptions occur. Customers...
Designing Chips In The Context Of Rapidly Evolving AI
Chip architects are grappling with the accelerating pace of AI model evolution, especially for edge‑centric, agentic workloads. Experts from Arm, Cadence, Rambus, Siemens EDA and others stress that memory hierarchy, data movement and reliable‑availability‑service (RAS) now dominate performance‑power‑area (PPA) trade‑offs....