Today's AI Pulse
Macron bets on AI with €75 bn SoftBank pledge to cement France’s tech future
French President Emmanuel Macron is securing a SoftBank pledge of up to €75 bn (≈$86 bn) for AI projects, branding the effort “Project Marengo.” The plan leverages France’s nuclear power to build low‑cost data centres and position the country as Europe’s AI hub.
Also developing:
By the numbers: Waymo acquires Apple’s Arizona Proving Ground for $220M
QumulusAI and the Shift From GPU Scarcity to GPU Efficiency
QumulusAI announced more than $124 million in three‑year GPU‑as‑a‑service subscriptions with Hyperbolic and another AI inference platform. The contracts cover 1,280 Nvidia Blackwell GPUs deployed on 160 Lenovo and Supermicro bare‑metal servers linked by Cisco Nexus networking. By right‑sizing CPUs, memory and storage around the GPUs, QumulusAI claims a roughly 20% reduction in inference costs versus conventional builds. The move signals a broader industry shift from hoarding GPUs to optimizing their continuous, cost‑effective use for inference workloads.
The Model Doesn’t Matter: Inside the Race to Be the AI Production Platform Filmmakers Want to Use
Artlist unveiled Artlist Studio, an AI‑driven video production platform that bundles leading generative models such as Seedance, Kling, Midjourney, Veo 3 and Sora into a filmmaker‑friendly workflow. The tool promises to replace manual prompt engineering with a director‑centric interface that mimics traditional...
Claude Fable 5 Is Anthropic’s Most Capable Public AI Model, and Will Hand Your Conversation to a Weaker Model the...
Anthropic unveiled Claude Fable 5, the first public model in its high‑performance Mythos family, priced at $10 per million input tokens and beating most published benchmarks. The model sits behind safety classifiers that automatically hand off biology, chemistry, cybersecurity or...
Context Compression Finally Works in Production: New Research Cuts LLM Input 16x without the Accuracy Hit
Researchers from several top universities and Lawrence Livermore National Laboratory unveiled Latent Context Language Models (LCLMs), an encoder‑decoder system that compresses input tokens before they reach the decoder. By shrinking context up to 16×, LCLMs delivered 8.8‑fold faster inference on...
Terraform MCP Server Is Now Generally Available
HashiCorp announced the general availability of Terraform MCP server for both HCP Terraform and Terraform Enterprise. The server lets AI assistants such as GitHub Copilot, IBM Bob and Claude Code interact directly with Terraform via the Model Context Protocol, automating...
Beyond the Stack Trace: Why AI Requires a New Debugging Paradigm
Integrating large language models (LLMs) into applications breaks the deterministic assumptions that underpin classic debugging techniques. Because LLMs generate probabilistic outputs, identical inputs can yield different results, and traditional tools like stack traces, breakpoints, and simple logs often fail to...
Ottimate Launches First MCP Integration Among AP Platforms
Ottimate unveiled the first Model Context Protocol (MCP) integration for accounts‑payable platforms, allowing finance teams to feed AP data directly into large language models such as Claude and ChatGPT. The feature, part of the 2026 Summer Release, promises to cut...
Seed Hijacking of LLM Sampling and Quantum Random Number Defense
Researchers expose SeedHijack, a backdoor that exploits deterministic pseudorandom number generators used in large language model (LLM) token sampling to inject attacker‑chosen tokens without altering model logits. In a 540‑trial benchmark, the attack hits 99.6% exact token injection on GPT‑2...
Google AI Overviews Legal Risks Raise New Enterprise Governance Questions
Google’s AI Overviews, which replace traditional search snippets with AI‑generated answers, are now facing their first legal challenges. A German court held Google responsible for an Overview that linked publishers to scams, while a Canadian lawsuit alleges a false criminal...
The Self-Driving Enterprise Is Already Here with Fred Laluyaux of Aera Technology
In this episode, Fred Laluyaux, co‑founder and CEO of Aera Technology, explains the concept of decision intelligence and how it enables a "self‑driving enterprise" where AI agents make and execute decisions at scale rather than merely offering recommendations. He outlines...
White House Discussions Are Weighing Giving CISA Mythos Access
Federal officials are weighing giving the Cybersecurity and Infrastructure Security Agency (CISA) access to Anthropic's high‑powered AI model Mythos to run public‑facing vulnerability scans across government networks. While CISA has not yet deployed Mythos, insiders say agency access is "imminent"...
Enterprise AI Wins by Owning Action, Not Model
Pattern: Enterprise AI is shifting from “best model” to “who controls context, identity, governance, and action.” @Microsoft, @Google, @ServiceNow, @Databricks, @Salesforce, and @AnthropicAI/@TCS are all moving there. That means the most important AI pattern in the latest announcements —...
A Logistics Company Designed an AI Tool Inspired by Its Supply-Chain Veteran COO. Meet Uncle Phil.
Lazer Logistics has launched Uncle Phil AI, an artificial‑intelligence coaching tool modeled after its veteran COO Phil Newsome. The system pulls data from telematics, in‑cab video, maintenance, labor and yard‑management workflows to give site managers real‑time operational guidance across its...
Concentrix and Napier AI Take on AML Compliance in ANZ
Global services firm Concentrix announced a partnership with RegTech specialist Napier AI to deliver an AI‑driven anti‑money laundering and sanctions‑screening solution across Australia and New Zealand. The joint offering leverages Napier’s Continuum platform, already used by over 100 institutions, to cut...
Janus Henderson Taps Anthropic for All-In Bet on AI
Janus Henderson, the $480 billion active manager, has partnered with Anthropic and Percepta to embed Claude AI across its operations. The firm is building two purpose‑built platforms—PRISM for client intelligence and LIBROS for research management—using proprietary data. Claude will also be...
4 Defining Trends in Agentic AI
At Data Summit 2026, Sanjeev Mohan identified four defining trends in agentic AI, stressing that business units now drive AI adoption rather than IT departments. He explained that effective agents require not only data but also context, intent, and semantic structures...
AI Agents in Healthcare: 3 Snowflake Summit Takeaways
Snowflake unveiled new CoWork capabilities at its 2026 Summit, adding Cortex Sense for 83%‑accurate answers across siloed healthcare data, an Async Agent API for continuous monitoring, and personalized, governed workflows that operate inside existing tools. The platform’s Horizon governance layer...
Honeywell AI Pilot Aids Coker Unit Operations at TotalEnergies Refinery
TotalEnergies teamed with Honeywell to pilot the Experion Operations Assistant AI on the delayed coker unit (DCU) at its Port Arthur, Texas refinery. The system predicts steam pressure disturbances 10‑18 minutes earlier and lifted prediction accuracy to 85%, while also...
AI Founder Declares the 'Era of Token-Maxxing Is Coming to an End'
Pylon CEO Marty Kausas announced that the era of unchecked AI token consumption—often called token‑maxxing—is ending as firms impose spend limits. He disclosed that Pylon’s Anthropic bill surged from $400,000 to $1.4 million annually after crossing a 150‑seat threshold, and he...
Anthropic’s Claude Fable 5 Draws Backlash From Cybersecurity Researchers Over Overzealous Guardrails
Anthropic released Claude Fable 5, its latest generative AI, but cybersecurity researchers quickly flagged its safety guardrails as overly restrictive. IBM X‑Force’s Valentina Palmiotti and Tolmo’s Matt Suiche reported that even basic tasks like code reviews or reading security blogs trigger blocks,...
What AI Benchmarks Miss About Real-World Performance
Enterprise AI teams are discovering that standard benchmarks overlook the latency, jitter and node‑degradation that occur in production. F5’s tests with MinIO show even modest network latency can cripple S3‑style object storage throughput, while jitter has a smaller effect. The...
Guidewire Cloud to Power Peel Mutual’s AI Strategy
Peel Mutual Insurance Company of Ontario has chosen Guidewire Cloud to accelerate its AI‑driven automation and modernise core operations. The insurer will deploy the platform across all lines of business with implementation support from NXT Level Technologies, a Guidewire PartnerConnect...
Nokia Adds Agentic AI to Its Network Services Platform
Nokia announced the integration of agentic AI into its Network Services Platform (NSP), allowing operators to deploy AI agents that reason over live network data and act within defined policies. The approach embeds AI directly in the platform that already...
How AI-Driven Reputation Challenges Are Reshaping Recruitment and HR Practices
AI-driven reputation damage is emerging as large language models reproduce outdated defamatory content, creating persistent digital harm for individuals and brands. The analysis outlines detection methods—direct output monitoring, training data tracing, sentiment drift—and recommends tools such as Brand24 ($69/month) and...
Barcelona-Based THEKER Raises €73 Million Series A to Accelerate AI Robotics Deployment
Barcelona‑based AI robotics startup THEKER announced a €73 million ($85 million) Series A round led by CRV, with participation from Samsung, LVMH, Henkel Ventures and other backers. The capital will speed up deployments with tier‑one industrial operators and expand the company’s software, electronics...
What the New AI-Powered Siri Really Means for Apple — and for OpenAI
Apple unveiled an AI‑powered version of Siri at its 2026 WWDC, slated for a fall iPhone rollout. The upgrade positions Siri as a direct substitute for the free tier of ChatGPT, leveraging on‑device data for faster, context‑aware answers. It marks...
Tilebox Launches Verifiable AI Workflows for Satellite Data
Tilebox announced a new infrastructure that enables verifiable AI workflows on Earth observation data, letting agents discover satellite imagery, execute analysis where the data resides, and generate immutable audit trails. The offering bundles a command‑line interface, agent skill set, and...
Market Brief: Goldman Says Investors May Be Underestimating AI Spending
Goldman Sachs warns that Wall Street is under‑estimating AI infrastructure spending, seeing demand for data centers, chips and power outpacing current forecasts. Oracle announced a massive AI and cloud capex plan of up to $95 billion for fiscal 2027 and will...
PTZOptics, LayerJot Partner on AI-Powered PTZ at InfoComm 2026
PTZOptics and AI startup LayerJot showcased a prompt‑based, robotic camera system at InfoComm 2026 in Las Vegas. The demo combined LayerJot’s Agentic Visual Application with PTZOptics SimplTrack3 cameras, all powered by Intel Core Ultra Series 3 processors for on‑device visual reasoning....
An Anthropic-OpenAI Price War Would Be Brutal
A fierce pricing battle has erupted between OpenAI and Anthropic, with both firms slashing rates for their enterprise AI models. OpenAI announced a 30% discount on its ChatGPT Enterprise tier, while Anthropic rolled out a new offering priced at roughly...
Why Validation Is Becoming the Proving Ground for AI in Life Sciences
AI adoption in life‑science IT is stalled despite broad generative AI use, with only 40% of firms seeing EBIT impact and most gains under 5%. Validation—documented proof that GxP software meets regulatory standards—offers a concrete, repeatable workflow where AI can...
Pleo Launches AI Agents for Autonomous Spend Management
Pleo, the European spend‑management platform, unveiled a suite of five AI agents that automate routine financial tasks from policy enforcement to treasury monitoring. The agents, including a live Policy Agent and an upcoming MCP agent, aim to shift finance leaders...
TCS and Anthropic Launch Global Premier Partnership to Drive Enterprise AI Scaling
Tata Consultancy Services (TCS) announced a Global Premier Partnership with Anthropic to bring the Claude family of AI models to enterprise customers. The collaboration creates a dedicated business unit, equips 50,000 TCS associates with Claude licensing, and targets regulated industries...
Employees Say They Spend More Time Managing AI than Producing Work
A Glean Work AI Index survey of 6,000 knowledge workers found AI saves about 11 hours per week, yet most of that time is spent managing the tools rather than producing output. Only 13% of respondents say AI has significantly...
Fraud Alert! AI Slop “Dream Girls” Are an Advertiser’s Nightmare
DoubleVerify uncovered a new AI‑driven fraud operation called DreamScheme, where scammers deploy AI‑generated women in provocative mobile ads to lure users to Made‑for‑Advertising (MFA) slop sites. The scheme routes clicks through a network of sites, including the hub TopZeno, generating...
DoorDash Launches AI Chatbot for Conversational Ordering
DoorDash unveiled Ask DoorDash, an AI‑powered conversational search that lets users find restaurants, meals and grocery items by typing natural‑language prompts. The tool can transform recipe links, cookbook photos or handwritten lists into a ready‑to‑order cart and even handle reservation...
Enterprise AI Agents Deploy in Hours, Not Weeks
This is a massive leap forward for enterprise AI. Parloa has completely transformed how AI agents connect to real business systems. What used to take weeks of custom engineering and middleware headaches can now be configured in hours by business...
Google Taps Samsung for Next‑gen AI Chip Amid Capacity Crunch
Google Turns to Samsung for Future AI Chip as Capacity Tightens — The Information https://t.co/HAr2qEUH6g
Grace Herman, Reveal: An Old Standard for New Tools: What Judges Expect From AI in Legal Work
Reveal surveyed California state and federal judges to gauge their stance on AI in legal work. The consensus shows judges are not hostile to technology but insist on defensibility and competency of AI tools. Lawyers who validate their AI applications...
AI-Powered Chat Saves Customers Lost at Checkout
Most online stores are losing customers in the chat box. Right before checkout. I explored with @gorgiasio how the fix isn't faster agents. It's a smarter AI. Here's the breakdown... https://t.co/OwRmXLs6CV
Healthee Unveils Zoe for HR, an AI Analytics Assistant for Benefits Performance
Healthee, the AI‑powered benefits navigation platform, launched Zoe for HR, an AI analytics assistant built into its HR suite. Zoe lets HR and benefits leaders ask plain‑language questions about engagement, utilization, and healthcare spend, delivering instant answers and proactive recommendations....
Robert Keeling & Ray Mangum: Results Are In: 5 Lessons From an Independent Study of aiR for Review
Redgrave LLP conducted an independent head‑to‑head test of Relativity aiR for Review, a generative‑AI‑driven workflow, against a traditional active‑learning review on a challenging set of roughly 45,000 documents. The documents required nuanced judgments about pharmaceutical marketing, controlled substances, and federal...
Keening, Mangum, Nelson, & Reiss: Better Than TAR. Nearly Expert: What a Major Study Shows About Gen AI and TAR...
A recent study comparing generative AI tools to technology‑assisted review (TAR) finds that AI can not only match but surpass TAR in complex document review. The research shows AI delivers higher recall, fewer false positives, and faster review cycles while...
Jeff Bezos Wants to Build an ‘Artificial General Engineer’
Jeff Bezos launched Prometheus, a startup aiming to create an "artificial general engineer" that automates engineering design and manufacturing. Backed by over $12 billion and valued at $29 billion, the company employs about 150 people and plans to apply chatbot‑style AI to...
Jeff Bezos Says AI Will Bring ‘Golden Ages’ Not Mass Job Losses
Jeff Bezos told investors that artificial intelligence will usher in a "golden age" of productivity rather than trigger widespread job losses. He highlighted Amazon’s aggressive rollout of generative AI tools across retail, cloud, and logistics, arguing that the technology creates...
More Than Hype: How DBAs Can Ensure AI Adds Real Value
Generative AI is rapidly becoming a core tool for database administrators, automating tasks such as SQL code generation, index design, and performance telemetry analysis. While the technology can slash manual effort and accelerate incident response, unchecked adoption raises compliance risks...
Storage Ticker - 11 June
Cohesity earned a USPTO patent for its generative‑AI backup platform, while Confluent’s new research reveals that 59% of UK executives still make major decisions by gut feel. CTERA launched a native n8n integration to securely expose enterprise file data to...
1 in 4 IT Leaders Say AI Mistakes Have Hurt Their Business
A new GoTo and Workplace Intelligence survey finds that one in four IT leaders say artificial‑intelligence mistakes have harmed their business, with 25% reporting impacts on customers or performance and 23% noting direct financial losses. Despite these setbacks, 90% of...
Shutterstock Evolves Its Content Library Into a Human-Led, AI-Powered Creative Platform
Shutterstock announced a revamped content library that functions as a human‑led, AI‑powered creative platform. The system blends its vast catalog of authentic, licensable assets with generative AI tools, conversational search, and proprietary Model Match technology. Contributors continue to earn royalties...
Headless or Walled: Cordial Draws the Line on How AI Agents Should Work
Cordial, the enterprise marketing platform behind brands like Levi’s and L.L.Bean, unveiled a headless, LLM‑agnostic infrastructure that exposes every core capability as a standard service. The new Model Context Protocol (MCP), CLI and API let both internal and external AI...