Så använder man nätverk

16/10/2026

An IBM news:

This week’s number: 700
That’s the estimated number of French supermarkets and food retailers affected by a widespread cyberattack targeting internet-connected refrigeration control systems used by technicians. Temperatures inside refrigerators reached as high as 50° Celsius (122° Fahrenheit) and have led to millions of dollars in spoiled food.

13/10/2026

IBM:

Custom GPTs: Attackers have a new way to implement an old trick

Imagine asking an LLM for help with a routine computer problem, only to be told to copy and run a command that secretly installs malware. That’s the trick behind a campaign recently uncovered by researchers at cybersecurity firm Huntress.

The attackers behind the incident used malicious custom GPT experiences to steer users into ClickFix workflows that deploy remote access trojan malware. ClickFix attacks typically disguise malicious instructions as a familiar task, such as completing a CAPTCHA, verifying a browser or following a troubleshooting step. This means the user copies and runs the command, effectively installing malware while the attacker waits patiently with their morning coffee.

“The general kill chain is still the same steps, but it’s very powerful,” said Norman Dorsch, Product Manager with IBM X-Force on the latest episode of Security Intelligence. “The user is the one actually installing the payload during that ClickFix attack.” A kill chain describes the sequence of an attack and, in this case, the human unknowingly takes part.

What makes the campaign so notable is that attackers do not need to compromise the AI model itself. Instead, they can exploit the credibility of the platform and the growing tendency to view all chatbot instructions as helpful, routine and safe. It’s a classic social engineering tactic: prey on human trust and instinct to get people to do something they otherwise wouldn’t.

As Dorsch explained, the new attack campaign works so well because trust does the heavy lifting. A prompt such as “prove you are human” can lead a user to paste a command into a terminal, execute the payload and establish command-and-control access.

The technique might be old, but it’s gotten a serious upgrade. ClickFix campaigns have been publicly documented for at least two years, according to Michelle Alvarez, Manager of IBM X-Force Threat Intelligence. She warned that the industry’s intense focus on frontier AI risks is overshadowing established threats that continue to evolve.

The campaign also points to a larger challenge for AI marketplaces. As custom assistants and agent ecosystems expand, malicious experiences may appear alongside legitimate tools, much as they have in app stores, browser extension repositories and other software marketplaces.

“Whenever a company has a marketplace of sorts, it is expected that it will have, at some point, a level of responsibility in terms of pruning out what is proven to be malicious,” said Giacomo Casoni, Senior X-Force Incident Response Consultant at IBM.

The experts’ takeaway for all defenders: AI security should extend beyond merely protecting models. It also means preparing users to question the instructions delivered through these trusted AI chatbots. Attackers don’t have to break the chatbot if they can continue counting on users to obey them.

13/10/2026

IBM: This is how an editor can use AI:

Carl Franzen sees a gap between what artificial intelligence can do and how businesses are putting it to use.

As executive editor of VentureBeat, a publication covering enterprise technology, Franzen has turned that conviction into working systems for his newsroom. In our continuing series (in the IBM Think newsletter)on how industry leaders use AI, Franzen explains how the technology is reshaping his reporting, writing and publishing.

At VentureBeat, he built a Slackbot using Anthropic’s Claude application programming interface (API) that monitors an editorial channel, sorts shared links into newsletter categories and writes summaries into a Google Doc.

“It runs on Google Cloud, and it’s the kind of internal tool I would previously have had to request from an engineering team,” Franzen told IBM Think in an interview.

Franzen also developed One Big Story, an automated news service that scans thousands of outlets, selects one story each day and uses multiple AI models to transform it into articles, videos and other formats. He said it costs just USD 5 to USD 10 a month to run automatically.

One of Franzen’s most potent AI tools are his colleagues—the people pressure testing AI’s technical capabilities and guiding which projects take hold.

“Let the most AI-savvy and hungry people in your organization experiment with whatever AI they want for work (within reason), and learn from them which lessons to take back to the wider teams,” he said.

His experience has also shaped his thinking about where companies should look for new technology, going beyond the industry’s established providers. “[T]here are tons of new startups and interesting new open-source models to choose from,” he said. For Franzen, this freedom of choice allows for more experimentation and faster learning.

Beyond newsrooms, Franzen points to routine business tasks such as processing invoices, tracking orders and turning dashboard data into presentations. “From what I’ve seen in my reporting, AI has made all of these tasks and processes far more efficient,” he said.

His enthusiasm for the technology comes with a qualification. He uses ChatGPT, Claude and the web data API Firecrawl to research and draft stories, then checks the sources and rewrites the material himself before publication.

“AI doesn’t do the work for me. The direction, reporting, judgment and the final edit are still mine,” Franzen said.

09/10/2026

IBM:

AI math advancements have reached yet another benchmark, with OpenAI publishing more than 700 manuscripts from an unreleased model just days after Meta shared several papers mathematicians wrote with its Muse Spark model.

Why it matters: AI is producing math research faster than humans can read it. Earlier conversations revolved around humans guiding AI and what we lose when it moves too fast. The latest calculation to solve for applies well beyond the world of proofs and theorems: what happens if we can’t keep up?

Driving the news: OpenAI’s math manuscripts cover 372 families of results, each of which the company says solves or makes substantial progress on an open problem in math or theoretical computer science. Many of the proofs have been checked by software for errors in their logic. Others haven’t, and OpenAI acknowledges those could contain mistakes.

What experts are saying: On this week’s episode of Mixture of Experts, the panel debated what these latest advancements in AI-generated math mean for the people who have to make sense of it all.

IBM Fellow Kush Varshney said math is more than the fractions and multiplication most of us learned in school. “It’s the thought process that goes into some world that’s defined by some axioms and postulates, and being able to definitively show with no doubt that some property holds,” he said. “It’s a journey of understanding.

Skyler Speakman, Senior Research Scientist at IBM, pointed out that human attention has limits. “We’re finite beings here,” he said. “Is this the beginning of a treadmill where humans [are] sorting out AI-generated content? We’ve seen that play out in other places, but is this where it really takes hold in the mathematics field?”

Between the lines: If AI is doing the work of finding the answers, humans carry the responsibility of making sense of them. Varshney cited a recent Quanta Magazine piece that raised the idea of mathematicians becoming interpreters. “It’s not the humans that are doing the math, but they’re interpreting it, they’re understanding it for the rest of humanity,” he said.

The other side: Utsav Dholakia, Forward Deployed Engineer at IBM, argued that humans falling behind might not be a loss, after all. “Is it really so bad if there is another tool, another sort of virtual intelligent being, that may help us uncover some fundamental aspects of the world around us?” he asked. “Maybe our brains are not even devised to think beyond three or four dimensions.”

What to watch: So far, AI has been tackling questions mathematicians have asked for decades. Speakman posed the next big one: “When will AI be able to ask the new math question?”

08/10/2026

IDC Research (IDC) introduced IDC Quanta for Tech Leaders, a new AI-powered technology intelligence platform designed to deliver fact-based, outcome-driven answers to the most pressing challenges technology leaders face today, backed by decades of IDC’s trusted research across major enterprise markets.

IDC Quanta for Tech Leaders brings together decision-grade answers sourced and cited from thousands of technology leader research documents, 15B+ proprietary data points tracked annually, vendor selection capabilities, and price benchmarking into a single AI-powered solution built specifically for CIOs, Heads of AI, CTOs, CISOs, and enterprise decision-makers. It works inside the tools leaders already use, so trusted intelligence is there to help them perform their jobs better and faster.

Data-backed intelligence to meet unprecedented challenges and seize new opportunities

Technology leaders face unprecedented pressure to navigate three simultaneous decisions with no established playbook: reevaluating their vendor ecosystem, controlling volatile AI costs, and proving measurable AI ROI. IDC research shows that 67% of enterprise AI agent budgets were overspent by more than 10 percent in the last 12 months, and 57 percent of CIOs don’t have a consistent, enterprise-wide vendor benchmark, leaving the majority exposed to unvalidated tech bets and compressed negotiation timelines.

“As AI transforms every corner of the enterprise, moving fast without validated proof creates an unacceptable level of risk,” said Lorenzo Larini, Chief Executive Officer of IDC. “IDC Quanta for Tech Leaders is a game changer for the industry. It puts actionable intelligence directly in the hands of technology leaders who in the past struggled with access that slowed down decision-making. Backed by decades of credible research, leaders can now confidently align strategy and investments to help their organizations make the right bets to succeed in this era of AI transformation.”

How always-on IDC intelligence drives results

Traditionally, the research industry has operated on a model of accessing large volumes of data through research documents and PDFs via a very small segment of individuals with access who curate the intelligence. IDC Quanta changes that model and aligns intelligence access with how work gets done today. It sits between the AI tools leaders already use—like email, Claude, and their enterprise AI tools—their own private data, and the full depth of IDC intelligence, delivering answers in the flow of work and contextualized for each leader.

IDC Quanta replaces fragmented, vendor-driven perspectives and unverified LLM outputs with independently benchmarked evidence that leaders can rely on. Now, leaders can:

Chart Strategies, Faster: Customers can continuously align technology roadmaps, vendor evaluations, and architecture decisions with independently verified market intelligence, IDC frameworks, and more than $500B in tracked tech market spend.
Win Their Mandate, Confidently: Tech leaders can bring the case for urgency, opportunity, and readiness with data-backed business cases formulated from benchmarks, cost models, vendor assessments, and vendor shares.
Prove Impact, Grounded in Data: Leaders can select the right technology at the best price, then continue optimizing costs and returns with IDC’s mature governance and ROI benchmark methodologies.
IDC Quanta for Tech Leaders draws on analyst research spanning thousands of technology vendors, thousands of pricing benchmarks, and more than 250 IDC MarketScapes every year. Central to this offering is IDC’s proprietary AI Economics research agenda, which will enable technology leaders to understand the true costs and business value of their AI investments, strengthening the foundation for their innovation strategies.

“We were investigating drones as first responder programs, and we didn’t really have a good sense of why it just wasn’t affordable,” said Andrew Shipe, Innovation Strategist for Sourcewell. “Working with IDC and [IDC] Quanta, we were able to put some actual numbers to the pain that these agencies were telling us about and really start to understand what the problem was.”

An adaptable platform designed for diverse enterprise needs

IDC Quanta for Tech Leaders brings IDC’s research and data, expert analysts, and hands-on deal support together for one predictable annual fee, not a percentage of savings. It is built around the three things tech leaders told IDC they value most: research and data they can trust, analysts they can call, and specialists who review the deals.

IDC Quanta for Tech Leaders will be available in seven packages at different price points and capabilities. Three entry-level packages make it simple to begin. One provides full access to IDC’s research and data. Another adds entry-level deal reviews, analyst inquiries, and price benchmarking. And a third is built for vendor selection and deal pricing reviews, including direct interactions with domain, specialized tech vendor, and market experts. Leaders start with the need in front of them and can add on as their decisions and investments grow. Premium packages add seats, additional analyst access, and tech leader event passes.

IDC Quanta is also now available on AWS Marketplace, allowing tech leaders to simplify and accelerate the procurement process, and apply existing AWS budget commitments (cloud credits) toward IDC spend. Customers benefit from the full offering and IDC experience, regardless of purchasing path.

“Technology leaders, whether leading technology in healthcare, retail, financial services, or manufacturing, must make critical decisions in lockstep with line-of-business partners who now control roughly half of all technology spend,” said Meredith Whalen, Chief Research and Data Intelligence Officer at IDC. “IDC Quanta for Tech Leaders eliminates friction by delivering vertical-specific benchmarks, industry-specific AI use cases, and tailored intelligence directly into the applications teams already trust. It moves analyst-grade insight out of portal silos and puts actionable, industry-specific proof right at leaders’ fingertips.”

Available now

IDC Quanta for Tech Leaders is available now. To learn more or to request a demonstration, visit https://www.idc.com/quanta/tech-leader/.

SIX DECADES OF RIGOR Full Access to IDC Intelligence and Experts. 120+ tech markets covered 1,000+ global experts 404,000+ tech leaders surveyed annually

05/10/2026

HPE:

It brings the NVIDIA Open Agent Safety Platform together with infrastructure-level security for private, regulated, and sovereign AI

Put Agent controls where they can be enforced

Autonomous AI agents do more than generate answers. They use tools, access data, write files, call APIs, and take actions across enterprise systems.

As their autonomy grows, enterprises need clear answers to three questions:

Where are our agents running? What can they reach? What did they do?

These are operational questions. Controls documented in a policy do not answer them. Controls enforced in the operating environment do.

NVIDIA Open Agent Safety Platform enhances agentic security at enterprise scale across the software that runs agents and the hardware and compute layers that power their work. NVIDIA OpenShell, the secure runtime at the center of the platform, is now generally available. HPE Private Cloud AI is integrating OpenShell to help enterprises apply consistent governance to agent ex*****on, with availability planned for Q4 2026.

Together, HPE and NVIDIA are addressing agent governance at two levels. The runtime governs what an agent can access and change. The underlying infrastructure helps protect data, network traffic, and the systems on which agents operate.

Govern what agents can access and change

OpenShell is open source software that places controls around the environment in which an agent operates. It uses isolated sandboxes and policy enforcement to govern what an agent can read, write, execute, and access over the network, including where inference requests go.

These controls operate outside the agent process. This separates the control boundary from the model, agent harness, or application it governs.

When an agent encounters a policy limit, it can request a change. Operators retain approval authority, allowing permissions to adapt without weakening the established boundary.

The general availability release adds multi-tenancy, Kubernetes-based high availability, extensibility, centralized observability, and policy automation. Enterprises can support different agents, models, and frameworks while applying a consistent runtime boundary across them.

OpenShell is also open source. For infrastructure that governs autonomous systems, transparency matters. Organizations can inspect, test, and improve the trust boundary on which they depend.

Bring governed agent ex*****on to HPE Private Cloud AI

HPE Private Cloud AI is a turnkey AI factory solution as part of the HPE AI Factory with NVIDIA portfolio. HPE is integrating OpenShell into an operating environment designed to connect agent ex*****on with enterprise identity, policy, approvals, observability, and audit.

This approach gives platform and security teams a consistent place to govern agent activity without requiring development teams to use one agent framework. OpenShell supports a variety of models, harnesses, and agents across cloud, hybrid, on-premises, and air-gapped infrastructure.

The objective is straightforward: give agents the access they need to complete approved work without giving them unrestricted access to hosts, networks, credentials, data, or model endpoints. IT Administrators control what the agents can access.

Development teams retain flexibility in how they build agents. IT gains a common boundary for controlling how those agents interact with enterprise resources.

Extend protection beyond the agent runtime

Runtime governance is one part of the architecture. Organizations also need infrastructure-level protection for data, network traffic, workload isolation, and operational evidence.

NVIDIA Sentry is an independent watchdog that monitors agent activity and enforces security policies out of band. It runs on NVIDIA BlueField-4 DPUs and uses NVIDIA DOCA for hardware-isolated monitoring, agent governance, and policy enforcement. NVIDIA DOCA on HPE AI Factory solutions provide software frameworks, libraries, and tools for infrastructure services such as policy enforcement, threat detection, and network protection. HPE plans to support NVIDIA BlueField-4 DPUs across the portfolio of HPE servers, AI rack-scale systems, HPE Private Cloud AI, HPE AI Factory at-scale, and HPE Sovereign AI Factory, with availability based on product lead times.

The NVIDIA OpenShell secure runtime boundary and agentic sandboxes runs on NVIDIA Vera CPU systems, including HPE ProLiant Compute DL394 Gen12, as well as other compute platforms. NVIDIA Sentry runs on BlueField-4 DPUs in a hardware-isolated security domain independent of the host software running those workloads.

HPE is also integrating NVIDIA Confidential Computing with HPE AI Factory solutions through HPE Services. NVIDIA Confidential Computing protects models and private data during ex*****on for on-premises and sovereign deployments. Cryptographic attestation and encryption establish a chain of trust that verifies hardware, software, and datasets. Availability is planned for Q4 2026.

These capabilities can help customers address requirements associated with the Cybersecurity Maturity Model Certification, the National Institute of Standards and Technology 800 series, Security Technical Implementation Guides, and Federal Information Processing Standards. The requirements addressed depend on the configuration and deployment.

Make sovereignty operational

For governments, public sector organizations, and regulated industries, agent governance raises additional questions: Who controls the system? Where does the data reside? Who is authorized to operate the environment?

A model or application cannot answer these questions alone. Organizations must address them across the infrastructure, operating model, and services supporting the AI system.

HPE Sovereign AI Factory extends the HPE AI Factory with NVIDIA portfolio for environments that require greater control over data, infrastructure, operations, and compliance. Combined with a governed agent runtime, this foundation can help organizations operate agents near sensitive data while maintaining defined access and operational boundaries.

Sovereignty is not only about where technology is located. It is about how infrastructure is deployed, controlled, secured, and operated.

Move agentic AI into production with evidence

NVIDIA Open Agent Safety Platform puts agent controls where they can be enforced, and OpenShell provides the governed runtime boundary while Sentry adds out-of-band monitoring and policy enforcement on BlueField-4. The HPE AI Factory with NVIDIA portfolio is integrating that boundary and connecting it to infrastructure designed for private, regulated, and sovereign operation.

This approach extends governance from the actions an agent takes to the infrastructure supporting those actions. It gives enterprises a clearer way to define access, enforce policy, observe activity, and maintain an auditable record.

Agentic AI is ready for production when an organization can show where its agents run, what they can reach, and what they did.

The next step is to make those answers enforceable, observable, and auditable.

Explore HPE AI Factory with NVIDIA

Scale production ready AI with secure infrastructure, software, and accelerated computing.

03/10/2026

IBM:

Asset management is stuck in the past. Can AI change that?
For years, asset managers have had a familiar answer when asked why they don’t modernize their maintenance operations: “Our data isn’t ready.”

Philippe Jetté, Product Manager for Asset Investment Planning at IBM, hears it frequently from capital planners in transportation, utilities, oil and gas, and other asset-intensive industries. Their organizations aren’t starting from scratch; they’re starting with something messier: disconnected systems, missing information and inconsistent data models.

The consequences are expensive. Reactive maintenance costs manufacturers USD 119 billion a year, while unplanned downtime can cost industrial companies anywhere from USD 10,000 to USD 500,000 per hour. It also takes a toll on equipment: in a 2020 NIST analysis, manufacturers relying on reactive maintenance saw 16 times the number of defects and triple the downtime of companies with a more modern approach.

While technologies like AI, Internet of Things (IoT) sensors, digital twins, and CMMS and EAM platforms all help, Vishal Rane, IBM’s Sustainable Facilities and Asset Management Product Manager, believes a more fundamental change is needed. “Instead of a break-fix approach,” he told IBM Think in an interview, “we should be thinking about everything we do for an asset, from inception and planning to procurement, operation, ex*****on and retirement.”

Rane believes that companies are leaving money on the table when they treat operations and maintenance as a cost center rather than a value engine. Instead of asking how to spend less, he and Jetté want them to ask a different question: “How much value can we find?”

Their answer is a three-part approach called Frame-Run-Refine, designed to connect maintenance decisions across an asset’s entire lifecycle more closely to capital planning and operations.

“Don’t wait for perfection,” Jetté advises. The companies that start with the assets, data and systems they have can build a business case for better data—and a much more sophisticated approach to asset management—in the future.

02/10/2026

IBM writes:

IBM Bob can now run inside a company’s own infrastructure, giving businesses a way to use AI coding tools without sending sensitive code to outside services.

In an announcement, IBM said customers could deploy Bob on-premises, in private and sovereign clouds and in air-gapped environments. The setup gives organizations more control over where their code and data remain while using Bob, IBM’s AI platform for developing, modernizing and maintaining software.

“For teams working with sensitive or regulated code, where the AI runs matters as much as what it can do,” John Duigenan, Director, Technology Sales Leader and Distinguished Engineer at IBM, told IBM Think in an interview. “Self-hosting Bob lets enterprises keep their source code and data inside their own environment, which makes it far easier to meet security, compliance and data residency requirements.”

Data residency and sovereignty requirements remain a challenge for many organizations. A June 2026 IBM Institute for Business Value study, The Calculus of AI Sovereignty, found that 68 percent of 1,000 senior executives surveyed said meeting those requirements across geographies was challenging. For companies operating across borders, the rules can determine where information is stored and which country’s laws govern it. That can complicate how businesses move and use data across different markets.

The infrastructure behind AI is also expected to become more distributed. Futurum Research projected that hybrid and edge deployments would account for 43.5 percent of the AI infrastructure market by 2030, up from 25 percent in 2024.

Customers using Bob’s self-hosted option can run supported models on their own infrastructure or connect Bob to supported external model services through hybrid configurations, according to IBM. The self-hosted option lets customers keep their code and data in their own systems while using Bob to build and update software.

For some development teams, that could determine whether they can use an AI coding tool at all.

“It turns Bob from a tool many of these teams couldn’t consider into one they can adopt with confidence,” Duigenan said.

01/10/2026

Claude Cowork, Claude Code and OpenClaw:

Commvault Protects and Recovers Agent Configurations and AI Data that Employees Can’t Afford to Lose on Their Laptops

As more employees rely on Claude Cowork, Claude Code, and OpenClaw to build agents on their managed devices, Commvault extends resilience capabilities to these AI tools and gives Security teams greater visibility into shadow AI

Commvault, a leader in unified resilience at enterprise scale, has announced that it’s enabling IT and Security teams to discover, protect, and recover agent configurations and other AI data in Claude Cowork by Anthropic, Claude Code by Anthropic, and OpenClaw (agentic AI tools) that employees are running on their managed Windows and macOS devices.

For workers and IT leaders, this can be a game-changing development. When a laptop is lost, damaged, replaced, or otherwise disrupted, organizations and employees – developers, marketers, and others – may not only lose files, but the agent configurations, accumulated AI context, and workflows they rely on to move critical projects forward. This could prove to be a devastating blow for employees and a productivity loss for organizations.

Via this new capability, IT teams can identify instances of certain agentic AI tools installed across managed devices, protect agent configurations stored locally, and recover that protected information if a disruption occurs. This also gives employees a recovery path for their valuable work – including AI-generated content, prompt libraries, custom instructions, and tool-specific context.

This capability is also useful for Security teams. It gives leaders a centralized view of supported AI tools installed across managed employee devices. If agentic AI tools are installed outside approved IT processes, teams can see where the tool is present and investigate. This gives organizations greater visibility into potential shadow AI.

This is well-timed. According to IBM, 80 percent of surveyed U.S. office workers reported using AI in their roles, while only 22 percent relied exclusively on AI tools provided by their employers. The result is a growing mix of approved AI usage and shadow AI, creating new visibility, protection, and recovery challenges for IT and Security teams.

“Most organizations cannot tell you which AI agents are running on their employees’ machines, let alone recover the work inside them,” said Pranay Ahlawat, Chief Technology and AI Officer, Commvault. “We give IT and Security visibility into where these AI tools are installed, and a recovery path for the context employees have built inside them.”

Global Availability

Discovery of OpenAI’s ChatGPT desktop app is available today. Discovery, protection, and recovery for supported local data associated with Claude Cowork, Claude Code, and OpenClaw will be generally available in the coming months. Customers can access these capabilities through Commvault’s existing endpoint protection offerings without a separate SKU or additional purchase. Support for additional AI tools is planned for future releases.

This offering is part of AgentRecover, Commvault’s solution for enterprise AI environment recoverability – from AI tools and their data and configurations to the connected systems they rely on.

Learn More in Nashville at SHIFT 2026

This announcement underscores the company’s commitment to helping organizations embrace AI while strengthening resilience. At SHIFT 2026, attendees will explore the new rules of resilience in the age of AI and learn how recovery, visibility, and cyber resilience can help organizations adopt AI with confidence.

SHIFT 2026 will be boldest in the company’s history, providing attendees with the opportunity to explore and immerse themselves in the new rules of resilience – and it all starts with AI.