Today's AI Pulse
Anthropic rolled out Claude Fable 5, a Mythos‑class model that blocks cybersecurity and biology queries via built‑in safeguards, routing restricted topics to Opus 4.8. At the same time, it introduced Mythos 5, a cyber‑capable version accessible only to roughly 200 vetted organizations through Project Glasswing.
Also developing:
By the numbers: Broadcom secures $35B debt financing for AI XPV Platform
Claude Mythos Turns N-Days Into N-Hours With Rapid Exploit Creation
Anthropic announced that its Claude Mythos Preview model can automatically generate working exploits for known vulnerabilities in minutes to hours. In tests, Mythos produced 16 functional exploits for Firefox and Windows, including eight Windows privilege‑escalation exploits within 18 hours. The model achieved its first working exploit in under one hour and did so at a cost of roughly $2,000 per Windows exploit. Anthropic warns that this accelerates the N‑day threat landscape and calls for “N‑hour” patching strategies.
Stay Ahead: Master Emerging AI Trends for Business
🚀 What’s next in AI? AI is evolving at an incredible pace, and the organisations that stay ahead of the curve will be the ones best positioned to innovate, compete and grow. I'm delighted to be partnering with LinkedIn Learning...
AI Coding Adoption Hits 97% but Governance Lags Behind
A recent UserEvidence survey of 831 engineers shows 97% of software teams now rely on AI coding assistants such as GitHub Copilot and Claude Code, yet only about 30% have formal governance policies. While 92% of respondents credit the tools with...
China Sets Out $295B AI Investment Plan
China is allocating roughly CNY2 trillion (about $295 billion) over the next five years to construct a nationwide network of AI‑focused data centres. State‑owned telecom operators China Mobile and China Telecom will manage the sites, while Huawei is expected to provide at...
AI Flags Breast Cancer Risk up to Ten Years Early
AI of radiologist read as normal mammograms shown to detect high-risk of breast cancer out several years "Artificial intelligence scores from sequential mammograms in individuals diagnosed with breast cancer showed elevated scores up to 10 years before diagnosis" a retrospective study, but supported...
Apple's New AI Explained: Gemini Distillation, EU Troubles and Third-Party Clouds
Apple announced a revamped Apple Intelligence suite built on five new foundational AI models co‑developed with Google. The partnership taps Google’s Gemini technology and runs part of the workload on Google Cloud, promising more personalized experiences across iOS, macOS, and...
Anthropic Says Mythos Is Too Dangerous for the Public. It Just Gave 150 More Organisations Access.
Anthropic warns that its Mythos AI model is too dangerous for public release, yet it has broadened access to roughly 200 organizations across 15 countries, adding 150 new participants in June. Mythos can uncover thousands of zero‑day vulnerabilities, chain them...
SpaceX Unveils 11-Million-Square-Foot Gigasat Factory, a New Manufacturing Facility for Space-Based Data Centers — Aims for 1 GW/Year of Space...
SpaceX announced an 11‑million‑square‑foot Gigasat factory in Bastrop, Texas, dedicated to mass‑producing AI‑focused satellites. The AI1 satellite, a 70‑meter orbital data center with a 150 kW compute payload, is slated for production start in 2027. SpaceX targets 1 GW of space‑based AI...
ZipRecruiter Expands AI-Powered Job Search with Claude Integration
ZipRecruiter has added a connector for Anthropic’s Claude AI assistant, letting job seekers search its marketplace through a visual chat interface. Users type preferences and receive real‑time, curated listings with salary, location and benefits displayed on easy‑read cards. The integration...
Will Douglas Heaven Forecasts AI Landscape in 2026
At SXSW London, our senior editor Will Douglas Heaven laid out how he’s thinking about AI in 2026.
Lighthouse Launches AI Assistant for Hotel Commercial Teams
Lighthouse unveiled Ernest, an AI‑driven assistant that unifies revenue, marketing, sales and distribution data for hotel commercial teams. The tool pulls information from over 500 hospitality platforms, allowing users to ask natural‑language questions and act on answers without hopping between...
Recursive Self-Improvement: The Latest From Anthropic
Anthropic’s recent essay frames recursive self‑improvement (RSI) as a concrete engineering loop where AI designs, tests, and deploys its successors. The company shows that Claude‑generated code grew from a few percent to over 80% of merged changes by May 2026, and...
Optimus Prime Will Take Your Call: Hasbro Leans Into AI-Driven Licensing
Hasbro has launched Sixth Wall, an AI studio designed to manage and license its character portfolio for generative‑AI applications. The studio introduces “behavioral licensing,” which safeguards a character’s voice, personality and interaction rules while allowing third‑party developers to use icons...
AI Data Centers Built in One Month, 70% Faster
Most AI data centers take a year or more to build. By the time they go live, the hardware roadmap has already moved on. @KAYTUS_ built its AI Compute Pod around that problem, and it now holds both the 2026 iF...
How to Use AI With Your Portfolio
The article shows how investors can upload a simple spreadsheet into an AI tool to get a quick portfolio review. By asking the AI to assess asset allocation and asset location, users receive textbook‑style feedback that highlights sector concentration, international...
Nebius Chooses Kao Data’s Harlow Campus for Major AI Infrastructure Deployment
AI specialist Kao Data announced a 22‑megawatt, ten‑year AI infrastructure deployment for Nebius at its Harlow campus in the UK. The agreement forms part of Nebius’s roughly $2.2 billion investment in the British market and will host the Nebius AI Cloud...
Beijing's $295 Billion AI Buildout Would Require 80 Percent Domestic Chips, Locking Out US Suppliers
Beijing is rolling out a 2 trillion‑yuan ($295 bn) five‑year plan to build a nationwide AI data‑center network, with state firms like China Mobile and China Telecom operating most sites. At least 80% of the hardware, including AI chips, must be sourced...
How to Build Agent Workflows in Your ATS
Recruiters are turning to AI agents inside Applicant Tracking Systems (ATS) to automate routine tasks such as candidate research, scoring, and outreach. The article outlines a step‑by‑step framework: start with a detailed Ideal Candidate Profile, connect the ATS to internal...
Keen AI, National Grid Secure Funding
Keen AI and National Grid secured a £355,985 (≈$455k) Alpha award from Ofgem’s Strategic Innovation Fund to develop the Foundational Shared Model Operations (FoSMo) AI system for electricity‑network asset management. The consortium—including SP Energy Networks, SSEN Transmission, UK Power Networks...
The Next Enterprise AI Wave Will Reach Beyond Agents, Says SAP Labs US Managing Director
At SAP Sapphire, Yaad Oren, SAP Labs US managing director, outlined six strategic technology areas the company is tracking beyond today’s AI agents: future AI architectures, advanced data platforms, next‑generation user experiences, robotics and physical AI, quantum computing, and evolving...
Why Apple’s A.I. Upgrade for Siri Won’t Be Available in Europe
Apple unveiled a next‑generation, AI‑powered Siri that can answer questions, automate tasks, and pull data from across a device and its apps. However, the rollout will exclude the 27‑nation European Union because the Digital Markets Act obliges Apple to open...
How to Make Sure You're Covered if Your Favorite AI Tool Bans You or Has a Major Outage (One Prompt...
Recent high‑profile outages at Cloudflare, AWS, OpenAI, and Anthropic have shown that reliance on a single AI provider can cripple core business functions for hours or days. The blog outlines how a chain of infrastructure providers and trust‑and‑safety automations can...
Hard Data Drives Safe Deployment of Physical AI
Leadership in Physical AI means going above and beyond traditional approaches to safety. We’ve refined and built tools @KodiakRobotics that allow us to deliver clear and compelling evidence of the Kodiak Driver’s safety. Hard numbers, not anecdotal evidence and assumptions. These tools...
Inside the Move From Generative AI to Agentic AI in Enterprise Finance
AT&T’s finance team is piloting an agentic AI workflow for manual journal entry preparation, using the LangGraph orchestration framework. The system separates repeatable data‑processing tasks from human judgment, embedding finance‑owned playbooks, node‑level audit evidence, and explicit approval checkpoints. Each workflow...
Festo Launches GripperAI Software to Simplify Flexible Robot Handling
Festo unveiled GripperAI, an AI‑driven software that lets industrial robots grasp mixed, unfamiliar items without extensive programming or dedicated vision systems. The platform automatically determines the best gripping point and selects the appropriate end‑of‑arm tool, whether vacuum or mechanical, in...
The Concerning, Unchecked Rise of E2E AI in Physical Applications
The article contrasts traditional deterministic engineering, exemplified by NASA’s Artemis II mission, with the rapidly expanding use of end‑to‑end, data‑driven AI for autonomous vehicles. It traces the paradigm back to Nvidia’s 2016 paper and Tesla’s Full Self‑Driving system, highlighting how raw...
GrubMarket Deploys AI Sales Agent to Automate Food Distribution Prospecting
GrubMarket unveiled a dedicated Sales AI Agent on its AI platform, designed to let wholesale food distributors discover prospects, analyze menus and generate price sheets in minutes. The tool aims to cut sales‑cycle time and boost revenue‑team efficiency across the...
AWS Rolls Out Agentic Shopping Assistant to Speed Retail AI Deployments
Amazon Web Services introduced the Agentic Shopping Assistant, a packaged AI solution that lets retailers build custom conversational shopping experiences in roughly 60 days. The service bundles architecture, starter code and expert support, aiming to slash development cycles from years...
Meta Mulls Multibillion‑Dollar Stock Sale, Launches $115M AI Workforce Academy
Meta Platforms is reportedly weighing a multibillion‑dollar secondary stock offering to fund its AI build‑out, a move that sent its shares down about 6% on June 5. At the same time, the company announced a $115 million America’s Workforce Academy to...
Microsoft Rolls Out Copilot to 505,000 NHS Staff, Saving 43 Minutes per Day per Worker
Microsoft announced that NHS England will extend Microsoft 365 Copilot to 505,000 clinicians and support staff after a trial showed an average 43‑minute daily productivity gain per employee. The rollout, slated for completion by October 2026, is positioned as a...
Apple’s Real AI Story Isn’t Siri: It’s a 20-Billion-Parameter Model that Runs From Your iPhone’s Flash
Apple unveiled its third‑generation Apple Foundation Models at WWDC, highlighting a 20‑billion‑parameter on‑device model called AFM 3 Core Advanced that runs directly from flash storage. The model uses a technique called Instruction‑Following Pruning to load only a subset of parameters into RAM, enabling...
Naver and Nvidia Team Up on Gigawatt‑Scale AI Factory, Led by Founders and CEOs
Naver and Nvidia agreed to build a gigawatt‑scale AI factory starting at Naver’s Sejong data centre, targeting 55 MW of capacity in early 2027 and scaling to 200 MW by 2028. The deal, highlighted by founder‑chairman Lee Hae‑jin and Nvidia CEO Jensen...
AI Maps Global Glacier Melt with Single Labeled Image
An AI model can now map melting glacier fronts around the world, with just one hand-labeled image per glacier and a few clever constraints. https://spectrum.ieee.org/tracking-glacier-melting-ai?share_id=9589874
19‑Year‑Old Founder Sells AI Calorie App to MyFitnessPal for $30 Million
Zach Yadegari, a 19‑year‑old entrepreneur, sold his AI‑driven calorie‑tracking startup Cal AI to MyFitnessPal for $30 million. The app, launched in May 2024, had amassed 10 million users and $30 million in annual revenue, highlighting the market appetite for AI‑powered health tools. Yadegari is...
Microsoft AI Chief Says Superintelligence Is Near, Warns of Data‑engineer Upheaval
Microsoft AI chief Mustafa Suleyman announced at the Build conference that the company is moving toward superintelligence, unveiling seven new multimodal models and a dedicated Superintelligence team. He warned that the scale of upcoming models will force a rethink of...
Tiger Data Launches PostgreSQL Extension Designed for AI Agents
Tiger Data, operating as Timescale Inc., launched Ghost, a managed PostgreSQL extension built for AI agents. The service offers instant, unlimited database forking and a copy‑on‑write storage layer that charges only for changed data. Pricing is based on active compute,...
New Platform Uses Cryptographic Invisibility to Protect AI-Built Applications
Atsign has unveiled AI Architect, a platform that embeds cryptographic invisibility into AI‑generated applications. The solution assigns a unique, non‑custodial cryptographic identity to every resource, ensuring authentication, authorization, and encryption by default. By making identities invisible to attackers, the platform...
Apple’s AI Pitch Will Live or Die by Its Privacy Promise
At WWDC 2026 Apple unveiled Apple Intelligence, a suite of AI features spanning iPhone, iPad, Mac, Apple Watch and Vision Pro. The company emphasizes that queries run on‑device whenever possible and otherwise use its Private Cloud Compute, which now relies on...
Hong Kong Showcases AI Applications for Construction at E&M I&T Day 2026
On June 9, 2026, Hong Kong’s Electrical and Mechanical Services Department staged the E&M I&T Day 2026 at the Hong Kong Science Park, gathering 26 exhibitors and over 400 participants from the Greater Bay Area. The event showcased AI-driven tools...
Meta Builds Tent Data Centres and Launches AI Creator Assistant as Facebook AI Push Accelerates
Meta is rapidly expanding its AI infrastructure by erecting six 125,000‑sq‑ft weather‑proof tent data centres in New Albany, Ohio, cutting build times in half. The sites run on 200 MW of modular gas turbines and are part of a $145 billion capital...
Snowflake Expands AI Offerings with New Development and Governance Tools
Snowflake announced a suite of AI development, governance, and data‑interoperability tools at its 2026 Summit, positioning its platform as the backbone of an "agentic enterprise" where AI agents act autonomously on trusted data. The updates include the expanded CoCo coding...
Taiwan Plans Stricter AI Chip Export Controls to China
Taiwan 🇹🇼 authorities are considering much stricter export controls on AI chip sales to China 🇨🇳 to further align with US 🇺🇸 measures - Bloomberg
As Enterprise AI Outpaces Governance, Alation Closes the Gap with New AI Governance Offering
Alation Inc. unveiled Alation AI Governance at the Gartner Data & Analytics Summit, offering a unified system of record for every AI model, agent, and tool. The platform automatically inventories assets, generates evidence‑backed model cards, and routes approvals through regulation‑aware...
Snowflake Expands AWS Collaboration with $6B Commitment to Accelerate Enterprise Agentic AI Adoption
Snowflake announced a multi‑year strategic collaboration with Amazon Web Services, committing $6 billion in infrastructure spend to accelerate enterprise agentic AI adoption. The deal deepens product integration, expands go‑to‑market via AWS Marketplace, and funds joint customer‑success programs. Snowflake’s Cortex AI will...
Rubrik Turns Its Platform Into an AI Agent and Ships Agent Cloud for Claude
Rubrik announced that its data security platform now operates as an autonomous AI agent, launching Rubrik AI across its Security Cloud and Agent Cloud suites. The company also made Agent Cloud for Anthropic's Claude generally available, featuring the SAGE governance...
Zscaler Launches AI Broker and Endpoint AI Security for AI Agents
Zscaler unveiled AI Broker, Endpoint AI Security, and AI Access Graph at its Zenith Live 2026 conference, extending the Zero Trust Exchange platform to protect autonomous AI agents. The AI Broker provides an agent registry and controls agent‑to‑agent and Model...
Researchers Build Self-Replicating AI Worm That Operates Entirely on Local, Open-Weight Models
University of Toronto researchers demonstrated a proof‑of‑concept AI worm that runs an open‑weight large language model on a single GPU to reason about targets, generate custom exploits and self‑replicate without human input or commercial AI APIs. In 15 trials on...
AI Tutor Tailors Curriculum, Revolutionizing Personalized Learning
i'm fully convinced this is the future of education Matt Pocock built a Claude skill called /teach, and the whole idea is a private tutor that builds an entire customized curriculum around YOU. think about how school works now. everyone gets the...
Apple Admits Siri's Future Runs on Google Gemini
Apple spent 15 years telling us Siri was the future. At WWDC on Monday, they finally revealed the future: it's powered by Google Gemini. The stock fell 5%. That's not a pivot. That's a confession.
Generative Vision Interview Questions #2 - The Isotropic Shortcut
In a senior AI engineer interview, candidates are asked to fetch the noisy state at diffusion step 500. While most write a 500‑iteration loop mirroring the Markov chain, production‑scale training on 100 M images demands a faster approach. The "Isotropic Shortcut"...