Today's AI Pulse
Anthropic shuts down flagship models after U.S. export‑control order
Anthropic abruptly disabled its Fable 5 and Mythos 5 models to comply with a U.S. export‑control directive, highlighting the fragility of businesses that rely on closed‑source AI services. The shutdown triggered a sharp rally in Chinese open‑source offerings such as MiniMax and Zhipu as investors bet on downloadable models that can run on‑premise.
Also developing:
By the numbers: SpaceX to acquire Cursor in $60B all‑stock deal
How AI Is Reshaping the Software Build Vs. Buy Decision
AI‑enabled development tools—from IDEs like Replit to agents such as Claude Code—are proliferating, allowing enterprises to accelerate custom software creation for UCC, ERP/SCM, and CX platforms. While AI can automate coding, testing, and documentation, studies show over 30% of AI‑generated code contains serious vulnerabilities, raising security and maintenance concerns. Off‑the‑shelf suites like Microsoft Teams, SAP ERP, and Zendesk provide pre‑built integrations and vendor‑managed updates, but may not fit unique workflows and can lock firms into a single provider. Organizations must weigh speed and innovation against governance, integration complexity, and long‑term cost when deciding to build or buy.
It’s Mythos’ World Now. How Do We Live in It?
Anthropic’s Claude Mythos, a large‑language model capable of uncovering massive numbers of zero‑day vulnerabilities, has been placed in a private testing program after startling results. The model’s ability to rapidly identify decades‑old flaws has spurred the formation of Project Glasswing,...
As Aerospace Prioritizes AI, Aviation Experts Ask, “What Problems Are We Trying to Solve?”
The AIAA Aviation Forum’s Forum 360 panel examined how AI can address aerospace’s core challenges of rising complexity, cost, and speed. Leaders from Lockheed Martin, Airbus, Boeing, Northrop Grumman, RTX and the Air Force Research Laboratory highlighted AI projects such as Lockheed’s Overwatch...
Ukraine’s Defense AI Chief Predicts ‘New Paradigm’ of Warfare
Ukraine’s defence ministry has turned artificial intelligence into a core combat tool, using AI to guide drones, analyse missile attacks and plan operations. Danylo Tsvok, head of the new AI centre, envisions a unified network that will recommend decisions from...
Kimi‑K2.7‑Code Boosts Performance, Cuts Reasoning Tokens
Moonshot AI's Kimi-K2.7-Code - Improved coding & agent performance over K2.6: +21.8% on Kimi Code Bench v2, +11.0% on Program Bench, and +31.5% on MLS Bench Lite. - Reasoning efficiency: Less overthinking, with 30% lower reasoning-token usage compared to K2.6. - Long-horizon coding:...
Tech Vendors Partner up to Address Enterprise AI Pain Points
Enterprises are hitting AI roadblocks around data readiness, governance, security and cost visibility, prompting a wave of vendor collaborations. IBM data shows most technology leaders feel under‑prepared for rapid AI deployment, while Gartner projects $2.5 trillion in global AI spending this...
Pairing Claude Code with Local Models
The article explains how Claude Code can be paired with local inference backends—Ollama, LM Studio, and llama.cpp—by redirecting its Anthropic Messages API calls via environment variables. It provides step‑by‑step installation, model download, and configuration instructions, highlighting models such as GLM‑4.7‑Flash,...
New Behaviour Analysis Reveals Deep Intent Differences Between ChatGPT Users and Google Searchers
BFJ Digital’s new behavioural report shows that ChatGPT is primarily used for early‑stage research, while Google Search continues to dominate the final purchase phase. The study finds users turn to conversational AI to synthesize product specs and compare options, then...
CSC and Sikt Partner to Advance Secure AI Across Nordic Public Sectors
Finnish research IT hub CSC and Norway’s public‑sector IT agency Sikt signed a Memorandum of Understanding to jointly develop secure AI services for education, research, culture and public administration. The agreement focuses on extending Sikt’s AI management platform, Sikt AI,...
Mistral AI Seeks 3 Billion Euros to Fund Its European AI Push
French AI startup Mistral AI is in early talks for a new funding round of about €3 billion (roughly $3.3 billion) that would value the company near €20 billion ($22 billion). The valuation marks a steep rise from the €11.7 billion ($12.8 billion) level recorded in...
The AI Companion Economy Is No Longer a Sideshow
The AI companion sector is emerging as a distinct consumer‑AI category, with 220 million app downloads and $82 million in H1 2025 spending, while global market size reached $14.1 billion in 2024. Europe’s AI adoption is accelerating, but the EU AI Act and GDPR...
EXL and Databricks Expand Collaboration to Help Enterprises Build Trusted Data Foundations for AI
EXL Service (NASDAQ: EXLS) announced it has achieved Gold Tier status in the Databricks Partner Program, deepening its collaboration to help enterprises build trusted data foundations for AI. Through its EXLdata.ai platform, the company integrates Databricks’ security, governance and lineage...
Bank Regulators Ramp up AI Scrutiny in Lending and Underwriting
U.S. banking regulators, led by the OCC and the Federal Reserve, have begun demanding detailed AI inventories during routine examinations, focusing on high‑risk functions such as lending, KYC and sanctions screening. The inquiries extend to vendor contracts, data protection, human...
Anthropic’s Claude Fable 5 Launch – What Matters to Legal
Anthropic unveiled Claude Fable 5, the newest Mythos‑class model that sits above the Opus line in capability and speed. The generative AI system delivers roughly double the throughput of its predecessor while maintaining higher factual accuracy. Fable 5 ships with...
TVBS Deploys AI Translation for NVIDIA GTC Taipei Keynote
TVBS deployed a self‑developed real‑time AI translation system for NVIDIA CEO Jensen Huang’s GTC Taipei 2026 keynote, delivering near‑simultaneous Chinese subtitles that many viewers thought were official. The in‑house solution was built and optimized in just nine days, a timeline...
MindWalk (NASDAQ: HYFT) CEO Dr. Jennifer Bath to Join Absci (NASDAQ: ABSI) and a Leading AI Compute Provider on Jones...
MindWalk Holdings CEO Dr. Jennifer Bath will appear on a virtual Jones AI Day panel alongside Absci and a leading AI compute provider on June 15, 2026. The discussion will focus on how high‑performance compute, native biological knowledge representations, and...
Untitled
Venti Technologies announced its AI‑powered autonomous vehicles have now traveled more than 500,000 miles and shifted over 340,000 shipping containers across global logistics hubs. The fleet, operating driver‑out AVs 24/7 for two and a half years, delivers 1‑inch positional accuracy...
Aymer AI Announces Venture to Develop 90-Acre Powered Land Bank in Texas
Aymer AI announced Project Macedonia, a venture to transform a 90‑acre Texas site into a powered land bank for AI and high‑performance computing. The development will provide up to 100 MW of hybrid grid and behind‑the‑meter power, enabling shovel‑ready parcels for...
Ep 797: Claude's Mythos and Fable 5, Google’s New Live AI, ChatGPT’s New Powers and 7 Other AI Features You...
In this episode Jordan Molson breaks down the week’s biggest AI releases, spotlighting Anthropic’s new Claude Fable 5 model, Google’s Gemini 3.5 Live translation, and a suite of under‑the‑radar updates like ChatGPT’s ability to send emails and generate interactive charts. He explains...
Why Google’s Exclusive Gemini Spark Is a Game Changer for Productivity
Google has rolled out Gemini Spark, an experimental AI assistant, to Workspace Ultra plan subscribers. The tool tightly integrates with Gmail, Calendar, Drive, and Docs to automate tasks such as email summarization, meeting‑brief creation, and file organization. Its customizable "skills"...
OpenAI Buys Another Firm for 24/7 Agents
OpenAI announced the acquisition of ONA, a platform that supplies secure cloud compute for AI agents to run continuously. ONA, originally a browser‑based development environment, now powers background agents that can operate 24/7 without user interaction. The deal signals OpenAI’s...
India Wont Be Second to Anybody in Developing Foundational AI Models: IT Secretary
India’s IT secretary announced that the country will not lag in building foundational AI models, unveiling Varya, the nation’s first video‑focused generative model from start‑up Avataar.ai. Backed by the government’s IndiaAI mission, Varya runs on 14 billion parameters and promises to...
Computex 2026: Aewin Showcases Next-Gen AI Infrastructure Platforms for Cybersecurity, Edge, and Data Centers
At Computex 2026, Aewin Technologies unveiled a new line of AI‑focused infrastructure powered by Intel Xeon 6 and AMD EPYC 9005 processors. The portfolio includes a 2U HA storage server with 24 NVMe drives, a high‑throughput network appliance, a dual‑socket general‑purpose server,...
ChatGPT Becomes Fastest Billion‑User App Amid AI Surge
Fastest app to reach the milestone. And the data shows that AI is a juggernaut (despite the public backlash) -> ChatGPT hits a billion monthly app users despite souring public AI sentiment "With its billion-MAU figure — achieved roughly 3.5 years...
TelecomTalk Live: How AI Is Helping Offline Retailers Compete in a Digital-First World
Offline retailers in India are turning to artificial intelligence to close the gap with digital‑first competitors. In a TelecomTalk Live interview, Zithara.AI co‑founders Varun Kashyap and Sri Devi Reddy explain how their platform leverages AI, behavioral segmentation and WhatsApp‑based conversational...
The AI Industry's Platform Trap Is Starting to Look a Lot Like Microsoft's
Anthropic unveiled Claude Fable 5, a Mythos‑based model, but secretly throttled its performance for tasks that enable customers to build competing AI products. After backlash, the company pledged to flag when a weaker model is served, though the restriction remains. The...
Generative Vision Interview Questions #5 - The Mode Ascent Trap
In a Midjourney senior AI engineer interview, candidates are asked why a perfectly converged score estimator still produces the same three images on every inference. The trap lies in using deterministic gradient ascent on the score function, which drives the...
How Beginners Are Using ChatGPT Codex to Automate Daily Workflows
ChatGPT Codex is being adopted by beginners to automate daily workflows, from drafting Gmail emails to generating visual content. The AI assistant interprets plain‑English prompts, allowing non‑technical users to create reusable “skills,” monitor token usage, and integrate with platforms like...
Freshworks Refresh 2026: Pivot to AI-Driven Employee Experience
Freshworks announced a strategic shift toward AI‑driven employee experience at its Refresh 2026 event in New York. The rollout includes AI Agent Studio, a no‑code builder, and the MCP Gateway that lets Freshservice connect to any third‑party AI tool. Executives said the...
OpenAI’s GPT-5.6 Release Is Triggering an AI Price War
OpenAI is preparing to launch GPT‑5.6, codenamed “Kindle,” which adds stronger reasoning, coding assistance, and vision processing. The company is signaling a possible price cut for the new model, igniting a price war with rivals such as Anthropic’s Claude Fable 5 and...
The Tech Download: Mistral's Arthur Mensch on Agentic AI, Chips and Enterprise Adoption
Mistral, the French AI startup that raised over $100 million, is positioning itself as Europe’s counterpart to OpenAI and Anthropic. CEO Arthur Mensch disclosed that the firm is not only expanding data centers powered by Nvidia GPUs but also exploring its...
Maneva Raises $27-Million USD Series A to Expand AI Vision in Manufacturing
Toronto‑based AI startup Maneva announced a $27 million USD Series A round, led by US Venture Partners with participation from Bling Capital, Freestyle Capital, Seguin Ventures, N49P and an SPV from Prem Kalevar. The funding brings the company’s total capital to $38.4 million...
Buying a Laptop May Soon Come with Instant Carbon Data, Thanks to New AI Agents
University of Washington researchers unveiled an AI‑driven system that can generate life‑cycle carbon estimates for laptops and other electronics in about a minute. The dual‑agent architecture—one analyst, one data‑gatherer—automates the data‑intensive steps of traditional LCAs, achieving a 5‑19% error rate...
ITWeb TV: AI Has Ushered in a New Work Order, Says Salesforce
Salesforce’s Africa lead Linda Saunders says AI is reshaping CRM, customer experience and the ICT channel, but many pilots still fail. The company insists on a clear value and cost case before launching AI projects, emphasizing a "system of context"...
The 127-Gigawatt Problem: Why AI Needs Its Own Power
The U.S. electricity grid is hitting a wall as AI‑driven data centers chase 127 GW of additional power, with a projected 28 GW shortfall for AI workloads alone by 2030. Hyperscalers are bypassing grid delays by funding on‑site generation, exemplified by Black...
Turn Text Prompts Into Fully Playable 3D Games
Tesana just killed every excuse for not shipping your game idea. They dropped Muranyi-3 and it's not just another AI game generator. This turns a plain text prompt into a fully playable 3D world with characters, physics, UI, and working game logic. Here's...
Fable 5 Lags Behind Opus 4.8 in Child Safety
Fable 5 is Anthropic’s worst-performing model for child safety. You can read the 2 example scenarios below to compare how Opus 4.8 and Fable 5 respond to the same situations. https://t.co/7abu7gndrK https://t.co/tKqUOrYCQK
DSIT Trials ‘AI Code Remediation’ to Help Power Ninefold Acceleration of Legacy Software Upgrades
The UK Department for Science, Innovation and Technology (DSIT) has completed early trials of AI code remediation tools to modernise legacy software. In partnership with Defra, the AI tools accelerated nine months of work into four weeks, enabling Defra to...
China Launches Photonics Lab to Power AI
Facing US chip curbs, China launches photonics lab to power AI with light Lab director says fundamental scientific challenges need to be overcome to achieve theoretical performance https://t.co/DyOqCSSOtV via @scmpnews
Speed Makes NVIDIA's LocateAnything Truly Actionable AI
Part 1: NVIDIA's LocateAnything is built for the moment AI stops answering questions and starts pointing, clicking, reading, and acting. Speed isn't a luxury — it's the difference between a useful agent and a confused one. https://t.co/kHFOKMDYOB https://t.co/lyEEc5ADUB
Tuya Smart Partners with Zeroth to Develop Robot-Driven Smart Living Experiences, Bringing Physical AI Into the Home
Tuya Smart and robotics firm Zeroth announced a partnership to fuse Tuya’s AI+IoT platform with Zeroth’s embodied intelligence, creating robot‑driven smart‑living experiences. The collaboration embeds Tuya’s Physical AI Engine and OmniMem long‑term memory into consumer‑grade home robots, turning them into...
MyndHaven Is Building AI-Powered Tools to Improve Therapy Preparation and Follow-Up
MyndHaven is developing an AI‑driven platform that captures daily emotional data from therapy clients and delivers structured weekly summaries to therapists, aiming to cut initial session time by over 90%. The system also generates post‑session recaps and offers employers anonymized,...
Prompt Injection Breaks Today’s AI Agents, Study Warns
Researchers from NTU, IBM, and other institutions released StakeBench, a benchmark that exposed systematic prompt‑injection failures in leading AI web agents powered by GPT‑5 and Gemini. Across 3,168 adversarial runs, indirect attacks succeeded 42‑68% of the time while direct attacks...
Microsoft Shares Slip as AI Cost Concerns Meet Xbox Restructuring
Microsoft’s shares fell 1.77% to $390.34 after reports of Xbox layoffs and heightened investor focus on the profitability of its AI‑driven cloud expansion. The market reaction underscores growing caution around large‑scale AI infrastructure spending, even as the company’s AI business...
MVP1 Ventures Rolls Out AI Agents‑as‑a‑Service to Push SaaS Beyond Chatbots
MVP1 Ventures announced its AI Agents‑as‑a‑Service platform, a structured offering that helps SaaS companies replace chatbot‑only interactions with end‑to‑end workflow automation. The service launches with three adoption plans and private‑cloud deployment options, aiming to turn AI experiments into measurable operational...
AI Is Cutting Hours of Office Work, but Also Creating a New Kind of Busywork
A new Work AI Institute survey of 6,000 digital workers shows AI tools save roughly 11 hours per week per employee, but also generate more than six hours of “botsitting” – monitoring and fixing AI output. While 75% of respondents report...
Guest Post: How AI Has Transformed Semiconductor Scheduling at NVIDIA and TSMC
Leading chipmaker NVIDIA’s artificial‑intelligence models are now embedded in TSMC’s fabrication plants, automating wafer‑level production scheduling. The AI evaluates millions of sequencing combinations in seconds, enabling instant rebalancing when tools fail or urgent orders arrive. Predictive dispatching and cross‑fab load...
Several Police Officers Arrested for Using Controversial Flock AI License Plate Reader System to Stalk Romantic Partners, Says Report —...
Investigations have uncovered at least 18 instances where police officers misused the Flock AI‑powered license‑plate reader (ALPR) system to stalk current or former romantic partners. In one highlighted case, an Orange City, Florida officer queried his ex‑girlfriend’s plates more than...
Child Safety Platform Rolls Out Across SA Schools
South Africa’s Online Safety Lab has teamed with US‑based ChildSafe.dev to launch RoseShield, an AI‑powered child‑safeguarding platform across schools and homes. The network‑level solution monitors all devices connected to a school’s internet gateway, shielding students from cyber‑bullying, sextortion, harmful content...
NVIDIA Confidential Computing Powers Apple Intelligence
NVIDIA’s Confidential Computing GPUs are now powering Apple’s Private Cloud Compute (PCC) for confidential inference, extending the workload from Apple‑owned data centres to Google Cloud. The deployment couples NVIDIA Blackwell GPUs with Intel CPUs that support TDX and Google’s Titan...