InformationWeek

AI's speed at identifying and exploiting vulnerabilities is turning the patching process on its head. The uptick in pace is forcing CIOs and CISOs to recalibrate the vulnerability identification and patching process without avoiding major disruption to enterprise operations. InformationWeek; Fernando Maymi, Anomali; Brian Wilson, SAS; Wally Dalrymple, ETS; Peter Bailey, Cisco; Atticus Tysen, NACD.DC, Intuit https://lnkd.in/eYtArj_i

Does AI cut through security alert noise, or just add to the pile? That's the question Clayton C. Peddy, CISO at ABBYY, and Erika Dean, CISO at Tricentis, dig into on the latest InformationWeek Podcast: "Does AI Help or Trigger Alert Fatigue?" They unpack what the alert cycle looked like before AI was introduced, whether AI improves the detection of false positives, and whether it is better to reduce the noise from alerts or to be hypervigilant with notifications for every potential security threat. 🎧 Listen here: https://hubs.li/Q04yL4xw0 A big thank you to InformationWeek and host Joao-Pierre Ruthh for a sharp, candid conversation. Learn about ABBYY’s AI privacy and security practices in the ABBYY Trust Center: https://hubs.li/Q04yL4rc0 How has AI changed alert fatigue for your team? #ABBYY #Cybersecurity #AIGovernance #AlertFatigue #CISO

InformationWeek asked me what Apple's new AI-ready Macs mean for CIOs. My comment: "I wouldn't have an Apple strategy. I'd have a workload strategy." Every workload has its own economics (and risks). On the desk: a developer's everyday coding help, or an analyst summarizing confidential files all day. Steady use, sensitive data, and fast response times make owning the hardware pay off. In your data center: a bank processing loan documents around the clock, or an insurer reviewing thousands of claims a day. High volume and regulated data justify shared capacity the company owns and controls. In the cloud: a marketing team testing a new idea, a year-end forecasting crunch, or any problem that needs the most powerful model available. Occasional, spiky, or cutting-edge work is cheaper to rent. This question about hardware architecture is increasingly capturing talk time in my client discussions. But the bigger thesis that keeps coming to mind for me is... When enterprise software moved to SaaS, the risk of getting value from the purchase shifted toward the vendor. Customers who didn't get value didn't renew. Entire customer success organizations grew up around that reality. Although I completely understand why leaders are buying their own AI hardware, many are shifting that business value risk back toward their enterprises without recognizing it. The vendor gets paid when the machine ships. Whether your people turn that compute into business value? That's on you. Put powerful AI machines on 500 desks. If people keep working the way they did before, the vendor still made its sale. And you now own an expensive deployment with negative return. Leaders need to get realistic with themselves...will your people actually work differently with that compute on their desk? And what if they DO work differently? Will that bring even more risk than you already face? What invisible risks lurk with all of that power, if your people don't know how to make good decisions in the way they use it? Picture an HR manager who uses a model on her new machine to screen 300 résumés in an afternoon instead of a week. She's working differently, exactly as leadership hoped. But the model quietly favors candidates who look like past hires, and she never checks why its shortlist looks the way it does. It's running on her desk, so no one else sees it either. Speed went up. So did legal exposure, and nobody knows it yet. When the technology is distributed, the accountability has to be, too. Every person using AI needs to own: 👉 What they should and shouldn't do with it 👉 What good output looks like 👉 When a human needs to step in Build that capability first. THEN buy the hardware. ---- Great article, Madeleine Streets. https://lnkd.in/g3WaycmG

In these week's news round-up: 💼 Oracle issues force majeure notice for Project Jupiter 🏗️ Akamai and Anthropic commit billions to AI infrastructure ⚖️ OpenAI calls for international AI technical standards. Or read on our website: https://lnkd.in/gh6VbuMS

For IT leaders, better observability means understanding how what’s happening across our systems is affecting the business, and knowing where to focus our attention. That visibility helps us make better decisions about where to invest, where problems are starting to surface, and when we need to step in before they have a broader impact. It’s becoming even more important as AI agents take on more work across the enterprise. Leaders will need a clear view into how those systems are performing and whether they’re operating as intended. Great perspective from InformationWeek on why observability is becoming an increasingly important part of the CIO’s job: https://lnkd.in/e88HM5up

In this week's news round-up: ⚖️ Microsoft proposes an AI code of conduct, ⚡ The AI Energy Management Alliance aims to expand data grid capacity, 🏛️ The National Archives briefly use Alibaba's Qwen model 🔓 Salesforce outage reveals dependency risks.

For IT leaders, better observability means understanding how what’s happening across our systems is affecting the business, and knowing where to focus our attention. That visibility helps us make better decisions about where to invest, where problems are starting to surface, and when we need to step in before they have a broader impact. It’s becoming even more important as AI agents take on more work across the enterprise. Leaders will need a clear view into how those systems are performing and whether they’re operating as intended. Great perspective from InformationWeek on why observability is becoming an increasingly important part of the CIO’s job: https://lnkd.in/e88HM5up

The Trust Problem CIOs Can't Ignore — and an Article That Nails It AI panic is real, and it's landing squarely on the desks of CIOs. When the leaders of Anthropic, OpenAI, Meta and Nvidia can't agree on how fast is too fast, employees, boards and customers are left absorbing the same alarming headlines — even though the AI most enterprises deploy is far simpler and far more contained. That's why this piece from InformationWeek is such an important read. Madeleine Streets, Senior Editor at InformationWeek, did a tremendous job attacking a genuinely tough subject: how do you build organizational confidence in AI when the technology's own creators disagree about the risks? She captured the nuance, brought in the right voices, and turned a messy industry debate into practical, actionable guidance. Bravo, Madeleine — this is exactly the kind of honest, CIO-lens journalism the enterprise world needs right now. A few thoughts I shared in the article: "A CIO can promise process: governed deployment, human oversight, continuous monitoring and transparency about what we know and don't know. What they shouldn't promise is outcomes... Boards don't need perfection; they need to know risks are identified, owned and managed. Promise diligence, guardrails and accountability — never certainty." "Say, 'Here's how we'll detect and respond to surprises,' rather than, 'There won't be surprises.'" And my favorite point from the piece: "Trust is built on visible restraint. The 'no' stories are your most valuable communications assets." Rejected proposals and pulled deployments are proof that AI has to earn its place in a workflow. Confidence in AI doesn't come from promises about the technology. It comes from demonstrated judgment. Read Madeleine's full article here: https://lnkd.in/ep4mrahT

An AI model can win the benchmark and still be the wrong choice for your business. The task, the data and the terms under which the model operates all matter.   Stephanie Overby’s latest InformationWeek article examines how companies are managing workflows that use several AI models, including who chooses the model and who owns the result, and I contributed a perspective on the governance decisions behind those choices. As I noted in the piece: “More sophisticated organizations treat model selection as fit for purpose, not best in show.” The challenge continues after selection. When work passes between models, separate logs can leave an organization with plenty of records and no clear account of what happened. A named owner, authority to pause the deployment and a record that follows the work across systems can help close that gap, with controls proportionate to the use case. Model changes also deserve fresh evaluation. A release announcement is an invitation to test, not an instruction to migrate. These are the practical questions we work through at Akerman LLP: how to give organizations greater flexibility to use AI while keeping responsibility clear. We explore that relationship between innovation and governance through Akerman Intelligence. Thank you to Stephanie for including my perspective alongside Yogaraj Jayaprakasam, Sumeet Mahajan, and Jeanine L Charlton. Read the full InformationWeek article here: https://lnkd.in/eenwRM6v