Still Little AI Return for UK Businesses, Accenture Survey Finds, and Many Reasons to Worry
Accenture’s "Generating Impact" survey reveals that UK executives remain skeptical about artificial intelligence, with 31% saying their operations would be unchanged if AI were removed tomorrow. Nearly half (46%) report little to no bottom‑line benefit from AI investments, and 25% cannot gauge the effectiveness of their AI spend. While senior leaders show tepid enthusiasm, employees are increasingly sourcing AI tools on their own, hinting at a disconnect between strategy and execution. The findings suggest many firms are underestimating how quickly consumers will adopt AI‑driven services, potentially skewing future budgeting decisions.
Atlassian to Train AI on Data From All but Its Top-Paying Users
Atlassian announced a new data‑collection policy that will automatically feed customer data into its AI training pipelines, except for its highest‑paying enterprise customers who can opt out. The change, effective August 17, 2026, applies to its flagship products such as...
GitHub Pauses Copilot Sign Ups, Tightens Limits as Agents Stretch Capacity
GitHub announced it is pausing new Copilot Pro subscriptions and imposing session and weekly token limits on existing accounts. The change follows a surge in usage driven by AI agents and sub‑agents that strain the service's compute capacity. GitHub is...
Inference Budgets Are Overrunning by "Orders of Magnitude" - What Now?
Goldman Sachs’ latest research shows AI inference costs are exploding, often exceeding original budgets by orders of magnitude. In a survey of 30 private and 10 public software firms, one company reported inference spend approaching 10% of its engineering headcount...
Europe Awards €180m in "Sovereign" Cloud Contracts, Touts SEAL Levels
The European Commission has awarded a €180 million (≈$156 million) contract to four cloud providers from Belgium, France, Germany and Luxembourg to deliver sovereign cloud services for EU institutions. The deal is part of a broader push for data‑sovereignty across the bloc,...
Palantir Doubles Down on Political Project and AI Weapons Amid Growing Concern From UK MPs
Palantir has renewed its push for a politically charged initiative, re‑issuing its founder Alex Karp's manifesto that advocates a more hawkish United States stance and the development of lethal AI systems. The move comes as UK Members of Parliament intensify...
STACKUP: The Stack's Weekly Tech Startups and Funding Wrap
Artemis, a cybersecurity startup focused on AI-driven defenses, announced a combined $70 million seed and Series A round as NASA’s Artemis II mission concludes. CEO Shachar Hirshberg discussed the growing AI security gap and the company’s solution. The weekly wrap also notes quantum‑AI...
There's Growing Pressure on the PyTorch Foundation. New Executive Director Mark Collier Tells Us His Priorities
Mark Collier has been named Executive Director of the PyTorch Foundation, taking over operational leadership as the AI ecosystem faces simultaneous shifts in hardware, software and model architectures. He brings experience scaling open‑source initiatives like OpenStack and will work with...
TSMC Says It Will Spend Close to $56bn to Keep up with Chip Demand
Taiwan Semiconductor Manufacturing Co. (TSMC) will invest close to $56 billion over the next two fiscal years to expand its production capacity and stay ahead of surging chip demand. The spending, the largest in the company’s history, targets new 3‑nanometer lines...
Digital Twins and Natural Disasters: The AI Tech Giving Emergency Response a Boost
Digital twins, long used for engineering simulations, are now being powered by AI to improve emergency response to floods, earthquakes and other natural disasters. In the past eight months researchers have demonstrated twin‑based planning in Turkey, South Korea and China,...
Mozilla Launches Open Source Thunderbolt to Shake up Enterprise AI Market
Mozilla announced Thunderbolt, an open‑source AI platform designed for enterprise workloads. The project aims to give companies full control over their AI stack, positioning it as an alternative to proprietary cloud services. Thunderbolt integrates with Mozilla’s existing developer ecosystem and...
UK Gov Announces First Startup to Get Funding From Its New Sovereign AI Unit
The UK government has announced the first startup to receive backing from its newly created Sovereign AI unit, a state‑run fund designed to operate like a venture‑capital firm. The initiative is intended to inject capital and strategic support into domestic...
Anthropic Releases Notably Better Coder Opus 4.7 – After Reports of Degraded 4.6 Performance
Anthropic has launched Opus 4.7, a new version of its AI coding model that promises a marked performance rebound after users reported a slowdown in the previous 4.6 release. Early tests show the model delivering higher accuracy and faster response times...
Critical MCP Vulnerability in Nginx-UI Now Actively Exploited in the Wild
The open‑source nginx‑UI, a web interface for managing Nginx configurations, has been found to lack authentication middleware, creating a critical Missing Control Plane (MCP) vulnerability. With over 11,000 GitHub stars and more than 430,000 Docker pulls, the tool is widely...
Overstretched NIST to Limit CVE Enrichments
The U.S. National Institute of Standards and Technology (NIST) announced it will stop enriching every CVE entry in its National Vulnerability Database due to a surge in submissions. CVE submissions rose 263 % between 2020 and 2025, overwhelming NIST’s resources. Going...
Shoe Retailer Allbirds Pivots to AI with $50m to Buy up GPUs
Allbirds, the sustainable shoe retailer, announced a $50 million investment to acquire approximately 1,660 NVIDIA H200 GPUs. The hardware spend marks a strategic pivot toward building in‑house artificial‑intelligence capabilities. By securing high‑performance GPUs, Allbirds aims to accelerate product design, demand forecasting,...
We Need to “Right-Size” AI Ambitions in UK, Microsoft VP Tells MPs
Microsoft’s UK vice‑president Hugh Milward testified before Parliament, warning that the country’s AI roadmap lacks clear direction and that the national power grid cannot support a 1‑gigawatt data centre today. He highlighted that without a reliable energy supply, ambitious AI...
Citi's Mandated Transformation Nears Its End, Bring on the AI Spending
Citigroup is scaling back its broad technology transformation program and trimming staff after reporting that 90% of its projects have hit or are close to their targets. The bank will redirect those resources toward a methodical, large‑scale rollout of artificial...
Patch Tuesday's a Monster: Thank AI?
Microsoft’s April Patch Tuesday delivered 247 patches covering 164 vulnerabilities, including eight critical flaws and two actively exploited zero‑days in SharePoint and Chromium. Security researcher Joe Desimone reported that all five of his local‑privilege‑escalation bugs were discovered using AI, highlighting...
Court Transcripts and Public Inquiry Responses: How the UK Gov Is Outsourcing Work to AI
The UK government is increasingly turning to artificial intelligence, notably Google’s Gemini model, to automate the drafting of court transcript summaries and responses to public inquiries. The Department for Transport has piloted Gemini for processing public consultation feedback, aiming to...
GitHub Finally Introduces Stacked Pull Requests, Devs Have Thoughts
GitHub has launched a private beta for stacked pull requests through its GitHub CLI, delivering a long‑awaited capability for developers. The feature enables a series of dependent changes to be submitted as a chain of pull requests, each building on...
Goldman Sachs Is "Accelerating" Cloud Investment for Better AI, CFO Says
Goldman Sachs announced an accelerated push to migrate its systems to the cloud, aiming to improve data quality and unlock AI‑driven productivity. CFO Denis Coleman said early results from the One Goldman Sachs 3.0 overhaul confirmed the need for stronger...
Axios Has a CVSS 10 Bug, Risks "Full Cloud Compromise"
The Axios HTTP client, downloaded over three billion times and embedded in roughly 80% of cloud and code environments, has been assigned a CVSS 10 rating under CVE‑2026‑40175. A proof‑of‑concept exploit shows the flaw can be escalated to remote code execution...
UK Police Spend £25m on Legacy Comms Network Waiting for New Emergency Network Service
UK police have allocated roughly $32 million to keep a legacy communications network running for six months, after a four‑year, $64 million contract for a new Emergency Network Service fell behind schedule. The original plan envisioned a seamless transition, but the new...
STACKUP: The Stack's Weekly Tech Startups and Funding Wrap
The Stack’s weekly roundup highlights a surge of AI‑centric funding across Europe and Asia‑Pacific, with the largest raises coming from those regions. Notable deals include a $100 million‑plus round for a data‑centre startup, a similar-sized investment in RISC‑V chip designer SiFive,...
Command Line: Mythos Burnout and the Boardroom
Anthropic’s Mythos large‑language model is prompting boardroom alarm as AI‑accelerated vulnerability discovery threatens to flood enterprises with exploitable findings. A new “Getting Mythos Ready” paper, authored by top CISOs, urges layered defenses—segmentation, canaries, honey tokens, and automated response playbooks—to contain...
Why IBM and Arm Are Partnering up on AI for Mainframes
IBM and Arm announced a strategic collaboration to develop dual‑architecture hardware that lets Arm‑based AI software run on IBM’s mainframe systems. The initial focus is on virtualization solutions that enable AI workloads to execute directly on on‑prem mainframes, bridging legacy...
Adobe Confirms Exploitation: Malware Uses Undocumented API
Adobe has confirmed active exploitation of the critical CVE‑2026‑34621 vulnerability in Acrobat and Reader on Windows and macOS. The flaw grants attackers arbitrary code execution when a user opens a crafted PDF. Exploit code abuses an undocumented API, SilentDocCenterLogin(), and...
France to Swap Windows for Linux Across Government as It Ramps up Sovereignty Drive
France announced a nationwide migration from Microsoft Windows to Linux for all government computers, targeting more than 100,000 workstations and thousands of servers. The initiative, overseen by the cybersecurity agency ANSSI, aims to cut IT expenses by roughly €2.5 bn (about...
Inside the Freshly Vendor-Neutral Payments Protocol Putting the 402 Status Code to Use
The Linux Foundation announced it will steward a newly created, vendor‑neutral payments protocol that operates directly at the HTTP layer. The protocol revives the rarely used HTTP 402 “Payment Required” status code to enable automated, machine‑to‑machine transactions for SaaS, API, and...
Container-Sized AI 'Pods' Could Be the Answer to Dragging Data Centre Plans, HPE Says
Hewlett Packard Enterprise (HPE) is mass‑producing container‑sized AI “pods” that function as miniature data centres, with a factory capable of delivering a unit in a matter of months. The modular pods bundle compute, storage, networking and cooling into a single...
France's OVH Answers Europe's Sovereignty Call with Dedicated Defence Unit
OVHcloud announced the creation of a dedicated defence unit aimed at serving European defence ministries and reinforcing data‑sovereignty initiatives. The new team will provide hardened cloud infrastructure, incident‑response capabilities, and compliance support tailored to government and critical‑infrastructure customers. OVHcloud’s move...
GOV.UK Eyes API Overhaul, GraphQL Could Do the Job
The UK government’s central digital platform, GOV.UK, is evaluating a major API redesign, exploring a shift from its legacy REST endpoints to a GraphQL‑based architecture. Officials say the change aims to deliver more flexible, robust and easier‑to‑integrate services for citizens...
Hugging Face's Safetensors, Meta's Helion Join PyTorch Foundation
The PyTorch Foundation announced that Meta’s Helion DSL and Hugging Face’s Safetensors format have joined its portfolio of open‑source AI projects. Helion provides a domain‑specific language for writing custom machine‑learning kernels, while Safetensors delivers a fast, secure tensor serialization method....
GitHub Admits It Needs "Deep Architectural Work" As Outages Persist
GitHub has experienced a series of high‑profile outages over the past few months, prompting senior leadership to acknowledge systemic reliability gaps. The company’s senior vice president of platform engineering said the platform requires "deep architectural work" to address scaling bottlenecks...
UK’s Biggest Police Force Gears up for £1 Billion ERP Overhaul, Swaps SI for SI, Oracle for Oracle
The Metropolitan Police has awarded DXC a contract worth up to £1 bn ($1.3 bn) to replace its aging ERP platform, P‑SOP, with Oracle Fusion SaaS. The legacy system, implemented in 2015, relies on 55 secondary interfaces and has already cost the...
Anthropic's Zero Day Machine "Mythos" Triggers Hype, Criticism
Anthropic announced Mythos, an unreleased frontier model it says can automatically uncover zero‑day software vulnerabilities far beyond its Opus 4.6 system. The company claims the model’s potency is so high it cannot be released publicly yet. Critics argue the assertions lack...
Google Makes It Easier for PyTorch Users to Switch to Its Own AI Chips
Google officially unveiled TorchTPU, a native PyTorch backend for its Tensor Processing Units, aiming to simplify migration from NVIDIA's CUDA ecosystem. Engineering lead Lee Howes highlighted the mature compiler stack and production‑grade reliability of TPUs as key advantages. The initiative...
UK Eyes Deeper “Sovereign” Space Surveillance Capabilities
The United Kingdom is advancing its sovereign space agenda by initiating a procurement process for a new space‑surveillance radar system, with a possible deployment in Antarctica. The effort aims to bolster the nation’s ability to track low‑Earth‑orbit objects and protect...
Trump Eyes $700m Cut for Already Depleted CISA
The White House, under President Trump, is proposing a $700 million cut to the Cybersecurity and Infrastructure Security Agency (CISA) budget, targeting its anti‑misinformation initiatives. CISA’s funding is already strained, with recent appropriations leaving the agency operating at a deficit. The...
Google's New Apache-Licensed Open Models Have Devs Excited
Google announced a suite of new large language models released under the Apache 2.0 license, marking a shift toward fully open‑source AI. The models deliver performance comparable to leading commercial offerings while allowing developers to run them on modest hardware....
Anthropic Signs Massive Deal with Google and Broadcom for More Compute as Devs Hit Limits
Anthropic has struck a multi‑year, multi‑billion‑dollar partnership with Google Cloud and Broadcom to dramatically expand its compute capacity. The deal will provision roughly 10,000 new TPUs and custom Broadcom AI chips, allowing the company to lift the recent restriction on...
Pyongyang, versus Nebraska?
North Korean state‑backed group UNC1069 infiltrated the popular Axios npm package, compromising two releases that were downloaded by millions of developers. Within three hours the malicious versions infected roughly 3% of cloud environments, according to cloud‑security firm Wiz. The breach...
STACKUP: The Stack's Weekly Tech Startups and Funding Wrap
The Stack’s weekly roundup spotlights a staggering $122 billion capital infusion into OpenAI, dwarfing even late‑stage Series G rounds. Smaller AI‑focused startups—AI SOC, orchestration platforms, and code‑review tools—also made headlines. Additionally, AI development platform Coder announced a $90 million Series C to accelerate growth....
MLPerf's New Inference Benchmarks Put NVIDIA on Top, But...
The MLPerf Inference v6.0 benchmark, released this month, evaluates AI inference workloads typical of modern data‑centre environments. NVIDIA’s Blackwell Ultra GPU emerged as the overall leader, topping the majority of the updated tests. AMD’s MI300X and Intel’s Gaudi 3 also posted...
"The Year of Surgical Refactors": $400 in Tokens Saves $500k in Annual Costs, Says Former Vibe-Code Sceptic
The article details how a new JSON query‑and‑transform language built in Go slashes latency and Kubernetes expenses. A modest $400 token purchase unlocked roughly $500,000 in annual cost savings, illustrating a high‑return refactor. The author, once skeptical of vibe‑code, now...
Kubernetes Is Finally Getting Rid of externalIPs
Kubernetes will remove the long‑standing externalIPs field in the upcoming v1.36 release, slated for the end of April. The change begins with deprecation warnings in the current version, giving operators time to adjust. ExternalIPs have been a frequent source of...
A Hitchhiker's Guide to RSAC: What You May Have Missed, From Post-Quantum to NSA Veterans
The RSA Conference (RSAC) highlighted a surge in post‑quantum cryptography initiatives, with several vendors unveiling prototype algorithms and migration roadmaps. Former NSA cyber experts presented new threat‑intel platforms that blend AI with traditional analytics, aiming to shorten detection cycles. Cloud‑native...
Google GAs New Ransomeware Protections for Drive, 14x Improvement From Beta
Google announced a new ransomware protection suite for Google Drive that leverages advanced AI to scan and block malicious files before they reach users. The service delivers a 14‑times improvement in detection rates compared with the previous beta program, while...
Nike Cuts Tech Roles, Continues IT Efficiency Drive After Finops Wins
Nike announced continued IT efficiency measures, cutting additional technology roles after a $230 million severance bill tied to recent supply‑chain and tech layoffs. The company dismissed its chief technology officer in January and placed the new chief operating officer in charge...