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
Insurance Denials Meet Their Match in AI-Powered Appeals
AI startups are reshaping the healthcare appeals landscape by automating denied‑claim letters and filing processes. Claimable, founded in 2023, uses a curated large‑language model to generate appeal letters for 28 conditions and 90 treatments, charging $50 per case. The company reports a 75% success rate and has secured $10 million in funding, including from Mark Cuban, while nonprofit Counterforce Health offers a free version. Insurers are also deploying AI, creating a three‑player arms race that could dramatically reduce the gap between patient entitlements and actual reimbursements.
3 Practical Ways AI Threat Detection Improves Enterprise Cyber Resilience
Enterprises are overwhelmed by thousands of low‑value security alerts, prompting a shift to AI‑driven threat detection. By building behavioral baselines for users, devices, identities and cloud workloads, AI reduces noise and surfaces genuine threats faster. Cross‑telemetry correlation and automated triage...
AI M&A Hits $155B, Nearly Half Involving Small and Mid‑Size Startups
AI‑focused mergers and acquisitions totaled $155 billion last year, and nearly half of that value involved small and mid‑size startups. San Francisco‑based boutique Axom Partners emerged as a key conduit for big‑tech buyers, underscoring a shift in private‑equity sourcing strategies toward high‑growth,...
Inside Freshfields' Partnership With Anthropic and What It Means for the Legal Industry
Freshfields Bruckhaus Deringer has entered a strategic partnership with AI pioneer Anthropic to co‑develop next‑generation generative AI tools for the legal sector. The collaboration gives Freshfields a seat at the table to shape Anthropic’s product roadmap, ensuring new models address...
Amazon and Walmart Push AI Deeper Into the Shopping Aisle
Amazon and Walmart are racing to dominate the emerging "decision layer" of retail, where AI‑driven recommendation engines, conversational agents, and dynamic pricing shape what shoppers see and buy. Walmart is pairing generative AI tools with its vast network of physical...
DTA Sounds Alarm on ‘Low‑effort’ AI
The Digital Transformation Agency (DTA) warned that the rush to adopt low‑effort AI tools could undermine inclusive system design and public‑service outcomes. Deputy CEO Lucy Poole urged civil servants to scrutinise AI solutions rather than rely on quick‑fix bots. She emphasized...
Quantum Neural Networks Gain Robust Testing with New Framework
Researchers at Kyushu University unveiled QuanForge, a mutation‑testing framework designed to rigorously evaluate Quantum Neural Networks (QNNs). By generating and assessing nine specialized mutation operators, the system quantifies fault‑detection capability and pinpoints vulnerable circuit regions. In simulated experiments, QuanForge achieved...
Tesla to Spend $2 B on AI‑Hardware Acquisition and Leverage Intel 14A for Terafab Chips
Tesla disclosed a $2 billion stock‑and‑equity deal to acquire an undisclosed AI‑hardware firm and confirmed it will build its next‑gen Terafab chips on Intel’s 14A node. The moves underscore Musk’s aggressive shift toward in‑house silicon to power autonomous vehicles, robotics and...
Opsera Launches Forge AI Software Factory to Cut Legacy Costs
Opsera announced Forge, an intent‑ and context‑aware AI software factory that claims to generate production‑grade applications in minutes. The platform aims to reduce the roughly 40% of enterprise IT spend tied up in legacy maintenance by embedding governance and compliance...
Stellium’s Hourly‑matched Renewable Power Slashes UK Data‑centre Emissions by 75%
Stellium Datacenters has moved to an hourly‑matched renewable electricity model for its Newcastle‑area high‑performance computing site, achieving a 95.4% renewable match and a 75% drop in carbon emissions. The strategy, built with Good Energy, is set to rise to 97‑98%...
Tesla’s $25 B AI and Robotics Bet Triggers Stock Volatility and Higher CapEx
Tesla disclosed a $25 billion capital‑expenditure plan for AI software, chips and robotics, sending its shares down nearly 3% after an initial rally. The spending surge, driven by the Terafab chip fab and Optimus robot production, raises questions about cash flow...
Don’t Mention the Government… Microsoft Splashes $25bn on .au .ai
Microsoft announced a $25 billion investment in Australia to build new digital infrastructure, boost national cyber‑defence capabilities, and fund workforce‑skilling programs. The figure, roughly half the projected $50 billion annual cost of the National Disability Insurance Scheme, underscores the scale of the...
Former U.S. Cyber Director Sounds the Alarm on Anthropic’s ‘Too Powerful’ AI Model
Anthropic’s new AI model Mythos can autonomously discover and exploit software vulnerabilities, prompting former U.S. Cyber Director Kemba Walden to label it “too powerful” for public release. The model has already identified thousands of high‑severity bugs across major operating systems...
Xbox Names Asha Sharma CEO, Slashes Game Pass Price and Rethinks Exclusivity and AI
Microsoft appointed Asha Sharma as Xbox chief executive and immediately cut Game Pass Ultimate to $22.99 a month, down from $29.99. In a joint memo with content chief Matt Booty, Sharma announced a reevaluation of exclusivity windows and generative‑AI use,...
Hiring ML Scientist to Shape Data for Advanced AI Agents
🚨Hiring Staff Machine Learning Research Scientist, Agents San Francisco, CA; Seattle, WA; New York, NY This role is at the intersection of cutting-edge AI research and practical application, with a focus on studying the data types essential for building state-of-the-art agents, such as...
OpenAI Targets Coding and Knowledge Work with Its New GPT-5.5 Model
OpenAI unveiled GPT‑5.5 and GPT‑5.5 Pro, promising faster response times and greater autonomy for coding and knowledge‑work tasks. The models claim to retain GPT‑5.4 per‑token latency while delivering higher intelligence and using fewer tokens on Codex‑style programming problems. They are rolling...
Databricks Opens Public Preview of Lakeflow Designer, a No‑Code AI‑Native Data Prep Tool
Databricks announced today that its Lakeflow Designer is entering public preview, offering a visual, no‑code, AI‑native environment for data preparation directly within the lakehouse. The launch targets analysts and domain experts, promising faster, governed workflows without moving data, and signals...
Can You Patent Your AI? New Guidance in Canada and the US for Software and AI Inventors
Corporate AI investment is spurring a surge in patent filings, prompting both Canada and the United States to clarify eligibility rules. The Canadian Intellectual Property Office issued a March 2026 Practice Notice that re‑emphasizes a physical‑effect test and provides a worked...
I Just Wanted Endpoints
The author highlights a missing orchestration layer—dubbed Layer 2C or the Reasoning Plane—between AI hardware and inference endpoints. On a single NVIDIA DGX Spark, they manually juggle vLLM and Ollama containers, deciding model placement, memory swaps, and runtime selection. At cloud...
The Spurs Cleared the AI Productivity Dip in Six Months
Enterprises adopting generative AI often encounter a short‑term productivity dip before realizing gains, a pattern described by Stanford economist Erik Brynjolfsson’s J‑curve. The blog notes that 94% of AI initiatives stall when firms neglect workflow redesign, governance, and data plumbing....
Oklo Shares Surge 9% After AI Partnership with NVIDIA and Los Alamos
Oklo Inc. (OKLO) climbed 9.12% to $79.01 on Thursday after unveiling a collaboration with NVIDIA and Los Alamos National Laboratory to fuse nuclear power with AI-driven modeling. The deal aims to create nuclear‑powered AI factories and accelerate fuel validation, positioning...
Google Launches AI‑Led Cyber‑Defense Agents at Cloud Next, Promising Minute‑Scale Threat Detection
At Cloud Next in Las Vegas, Google announced three new AI agents that automate threat hunting, detection engineering and third‑party context enrichment. The rollout expands an existing AI stack that already processes millions of alerts, shrinking manual analysis from half...
Zillow CEO Bullish About AI’s Potential, Opportunities for Agents
Zillow CEO Jeremy Wacksman told real‑estate leaders that AI is reshaping the industry. Internally, AI tools have already nudged employee productivity higher, and externally the company is rolling out features that automate note‑taking, generate floor plans and streamline marketing for...
Meta Turns Employees Into Data Source for AI Replacements
Meta is recording employee keystrokes, mouse movements, and screenshots across Gmail, LinkedIn, VSCode, Wikipedia. Stated purpose? training data for computer-use AI agents. Translation? the white collar workforce is now the dataset for its own replacements. Passwords, product roadmaps, immigration status, health info, all...
Why AI Shouldn't Manage Money Without Enterprise Safeguards
My friend @EvanKirstel is talking about something I have been feeling for a long time, and the reason why I have backed off from some agentic tools. I love putting AI in almost everything I do. I have it open on...
Nonprofit Launches Forum to Help City Leaders Leverage AI Effectively
The National League of Cities (NLC) has launched an AI and Emerging Technology Forum to help municipal leaders explore and implement artificial intelligence in local government services. The inaugural cohort includes 20 mayors, chief information officers and technology practitioners, with...
OpenAI Unveils Free ‘ChatGPT For Clinicians’ To Aid In Clinical Tasks
OpenAI announced on April 22, 2026 a new version of its large‑language model called ChatGPT for Clinicians. The service is offered at no cost to any U.S.-verified physician, nurse practitioner, physician assistant, or pharmacist. It is designed to help clinicians...
Hallucinations Are Different for eDiscovery Solutions. Here’s Why: EDiscovery Best Practices
The legal community is reacting to recent AI‑generated hallucinations, especially fabricated case citations that appeared in a high‑profile filing. While public large language models can invent facts, eDiscovery solutions operate on a fixed evidence corpus, so their errors are misinterpretations...
Meta Rolls Out AI Business Assistant To All Advertisers, Agencies
Meta announced the global rollout of its AI Business Assistant, extending the beta to advertisers and agencies of all sizes across major markets and languages. The tool, first piloted with U.S. small businesses, delivers real‑time, data‑driven recommendations, account restoration, spend‑limit...
Powerful AI Models Learn to Self‑Jailbreak Automatically
I mentioned this project in this week's AI Lab newsletter. It suggests a kind of flip side to the idea of self-improvement in AI models. As models become more powerful, it shows they can automate the process of figuring out...
Switching to API Pricing Inflates Bill Despite Unchanged Usage
POV: Your bill after getting forced to switch your OpenClaw AI agent from the $20 monthly subscription plan to the API pricing even though your usage stayed the exact same. https://t.co/q3e6xaquCn
Robots Learn to Feel What Vision Misses
Robots traditionally depend on cameras, but visual data degrades in low light, focus loss, or occlusion. A joint effort by Yonsei University and USC introduced a system that couples an RGB‑Depth camera with an inkjet‑printed tactile sensor array. The tactile...
AIBOM Not a Fix for Shadow AI, Says Critic
Subtweet alert... Hi @grok .. why do some people think that AIBOM is a solution to shadow AI? This is not accurate, right?
Six Pillars for Collaborative Robot Learning with Humans
A new Science #Robotics Viewpoint outlines six design pillars that could cultivate collaborative intelligence to enable robotic foundation models to learn with humans and other robots, and not just for them. @rienerr @s_albrecht https://t.co/9XG0blNqsK https://t.co/oqF0Vo1GwT
Reveel Offers AI-Powered Shipping Intelligence for Shippers
Reveel has launched Reveel IQ, an AI‑powered shipping intelligence platform that lets shippers query invoice and shipment data in plain English. The solution connects directly to carrier accounts, ingesting package‑level data to provide real‑time modeling, scenario simulation, and transparent cost breakdowns....
GPT‑5.5 Pro: Best Model Yet for Complex Tasks
Here’s my view on GPT-5.5, which I have been testing for a couple of weeks. It conducted not-bad social science research on its own, developed a novel RPG & more. There is still jaggedness but GPT-5.5 Pro is (for today) the...
Synthetic Data Promises Massive Leap over Fuzzing DBs
We've dabbled with agents and models using large fuzzing DBs as context. It was good. We are moving to tuning and training with synthetic data soon. Based on some experiments its gonna be epic. Synthetic data generated on private methodology...
Cloneable Secures $4.6M Seed Round
Cloneable, an AI automation startup focused on infrastructure operations, announced a $4.6 million seed round. The round was led by Congruent Ventures with participation from First In, Overline, St. Elmo Venture Capital, and Bull City Venture Partners. In the same announcement, Cloneable...
AI Excels at Editing, Not Original Writing
I just spent 6 hours writing a newsletter and 30 seconds using an LLM to chop off 4k words. AI is a crappy writer, but a pretty decent editor.
From AI Curiosity to Implementation: How Firms Should Evolve Their Strategy
John O’Connell, CEO of the Oasis Group, says wealth‑management firms are moving from AI curiosity to full‑scale implementation. He notes that most advisors are already experimenting, but the next step is operationalizing AI through governance, use‑case selection and policy. Oasis...
Social Recruiting Ad Company Acquired
GAIA, a European HR‑tech provider, announced the acquisition of SocialJobs, a social recruiting ad platform, to fast‑track its expansion into the United States. The deal gives GAIA access to SocialJobs’ AI‑driven job‑matching engine and roughly 1.2 million monthly active users. At...
Linux May Drop Legacy Network Drivers Amid Surge in AI-Generated Bug Reports
Linux kernel maintainer Andrew Lunn has submitted patches to strip out legacy ISA and PCMCIA Ethernet drivers, removing over 27,000 lines of code that support hardware from the 1990s. The move is driven by a surge in AI‑generated bug reports...
WARC: AI Redefines 'Companionship'
The World Advertising Research Center’s 2026 Consumer Trends report reveals AI is reshaping companionship, with 10% of global consumers reporting a relationship with an AI chatbot and 62% preferring AI for personal advice. The study links this shift to reduced...
OpenAI Unveils Its New, More Powerful GPT-5.5 Model
OpenAI announced the launch of GPT‑5.5, its latest flagship model, and made it instantly accessible to the hundreds of millions of ChatGPT users. The new system delivers notable gains in code generation and a broader range of office‑task capabilities. Unlike...
White House Accuses China of Industrial-Scale AI Model Distillation, Commits to Intelligence Sharing with OpenAI, Anthropic, Google
The White House Office of Science and Technology Policy issued a memo accusing China of running "industrial‑scale" AI model distillation campaigns that siphon capabilities from U.S. frontier models. OpenAI and Anthropic have documented millions of illicit queries and tens of...
VLLM-Lens: Fast Interpretability Tooling That Scales to Trillion-Parameter Models
vLLM‑Lens is an MIT‑licensed vLLM plugin that brings top‑down interpretability tools—probes, steering, and activation oracles—to trillion‑parameter models. Benchmarks show it runs 8‑44× faster than HF Transformers, nnsight, and TransformerLens on a single GPU, while supporting pipeline, tensor, expert and data...
Google Photos AI Enhances Old Pics Instantly
The "AI Enhance" feature in Google Photos on Pixel phones can dramatically improve even old photos with just a tap or two. Here's how to get started with this powerful new tech... https://t.co/WVe3OXhFyo #photography #android https://t.co/0JESADNrbz
The AI Journey Isn’t a ‘Sprint,’ Trintech CFO Says
Trintech CFO Omar Choucair says AI adoption is now an expectation for finance leaders, but it must be reliable, auditable, and aligned with the CFO’s zero‑tolerance for error. Trintech has embedded AI across its suite, launching the Beacon assistant to automate...
Diligent Debuts AI Board Member, Agentic GRC Workforce at Elevate 2026
Diligent unveiled AI Board Member, a secure AI assistant for directors, and a network of autonomous agents embedded in its Diligent One Platform at Elevate 2026. The AI Board Member can instantly recall board materials, industry news, and provide specialist perspectives,...
TikTok Tests AI Remix Option; What It Is, How To Opt Out
DirtyRoulette is an adult webcam platform that directly markets sexual encounters, operating as a front‑end for Flingster’s technology. Users can join for free and optionally upgrade to a VIP tier, with pricing starting at $7.99 for a trial week and...