Activity
15K followers
-
Ofer Klein shared thisNow more than ever, we need AI-powered vulnerability discovery for third-party apps, and here is why: Your ecosystem changes faster than a human team can review it. Each new app or agent adds to the problem. A fresh API or agent arrives with permissions of its own, so the surface your team is mapping grows while they are still mapping it. Manual testing shows the first strain. Privileges and settings change during the day, so an assessment you ran this morning describes a system that looks different by afternoon. A review that studies one app on its own also misses the route that crosses into another. So security teams give the first pass to AI. It reviews identities and permissions without pausing and keeps that view current while the environment moves. People still decide what to fix. AI gives them a picture current enough to decide on time.
-
Ofer Klein shared thisEvery security team needs to do this right now. The practices are ones your team already knows. What changed is the speed they run at, and the ground they cover. Start with what you can see. Build an inventory of every app and agent connected to your identities, so no access path stays hidden. Then configuration. A single weak setting looks harmless alone, yet combined with another exposure it becomes a step in an attack path. Remediate the over-broad permissions and loose sharing before someone chains them. Give your non-human identities the same care as your people. A service account reaches the same sensitive systems a human user does, so apply the ownership and least-privilege controls you already trust for staff. Then close the distance on detection. Automated attacks move on their own clock, so shorten the time between noticing suspicious activity and responding, with continuous monitoring and behavioral analytics. Your team already knows these moves. Run them continuously across the whole surface, and you can keep expanding what your AI does without losing sight of it.
-
Ofer Klein shared thisGartner projects that by 2028, 25% of enterprise generative AI applications will experience at least five minor security incidents per year. That’s up from 9% in 2025. The bigger concern is where much of that risk is coming from. Identity security was built around people: employees, contractors, joiners, movers, and leavers. Agents do not follow that lifecycle. They connect to SaaS platforms, call APIs, collect OAuth tokens, and gain permissions quietly. Because they never formally join or leave, access reviews can easily miss them. The result is predictable: unclear ownership, excessive access, weak audit trails, and more paths to sensitive data. Agents must be governed like any other privileged identity. They need a named owner, a defined business purpose, least-privilege access, regular reviews, and complete activity logs. The next identity security crisis may not begin with a compromised employee. It may begin with an agent no one remembered to govern.
-
Ofer Klein shared thisWe call it the 2am test. An AI system starts doing something it shouldn’t at 2am. One person needs to shut it down on their own. If that call takes three people to agree first, no one can take ownership of the AI risk. Passing this test takes four things. 1. The owner has to see what’s running. AI arrives through corporate cards and connectors approved in thirty seconds, so an owner accountable for a surface they have never seen is a hostage. 2. Their authority has to hold. Saying no is easy. The question is whether that no survives a business unit that wants the tool this quarter. 3. They need budget. Findings without funding simply document the risk instead of shrinking it. 4. There needs to be a consequence. And a failure has to cost them something. Score whoever is accountable for AI risk where you work, one point per condition.
-
Ofer Klein shared thisThis is what AI-scale discovery reveals about SaaS risk: Almost none of it is a software bug. The findings are the exposures a code scanner reads straight past, like a setting left open, or a trust link between two apps that no one approved. 𝗣𝗲𝗿𝗺𝗶𝘀𝘀𝗶𝝾𝗻𝘀 are where it starts. A user or an agent collects more access than the work calls for, and every unused grant gives an attacker another way to reach your data. 𝗖𝝾𝗻𝗳𝗶𝗴𝘂𝗿𝗮𝘁𝗶𝝾𝗻 comes next. Security settings slide off the baseline you approved, and each change that goes unnoticed widens the surface while your compliance posture reads stronger than it is. 𝝖𝘂𝘁𝗵𝗲𝗻𝘁𝗶𝗰𝗮𝘁𝗶𝝾𝗻 comes into play when an over-permissioned OAuth grant or an exposed token hands an attacker a working account, and through it, your sensitive data. 𝗦𝗵𝗮𝗱𝝾𝘄 𝝖𝝞 adds what you can’t see at all. An unapproved tool runs outside your governance, its data moving through routes no one watches. The finding that carries the weight is where these link together. A chain of permissions and trust across several apps becomes a route for lateral movement, and one small exposure turns into a wide compromise. Seeing every one of these at once turns a pile of settings into a clear view of your risk. That view lets your team walk the dangerous chain before an attacker does.
-
Ofer Klein shared thisWho owns AI risk? Ask five executives this question and you'll get back five different answers. Security points to Legal for the regulatory side, and Legal points right back for the technical one. The team that shipped the tool assumed someone signed off. All these answers are honest, which is exactly why it is so dangerous. AI risk never got handed to the wrong person. It never got handed to anyone in the first place. It settled into the space between the teams closest to it, each one assuming a neighbor had it. The org chart shows tidy boxes and hides the seams between them, and the seams are where the risk lives. The question gets asked out loud for the first time, on an incident bridge, at an hour nobody wanted to be awake. By then it is about blame, not process. The fix is having one person who can see what AI is running, kill it, and answer for it when it fails.
-
Ofer Klein shared thisA company who works with us told us everything changed the moment this happened: James Bowie, who runs security at Tampa General Hospital, put a single dashboard on screen in front of about a thousand of his company's leaders. It listed every tool each department had brought in, with the exposure attached: around $𝟏𝟓𝟎 𝗺𝗶𝗹𝗹𝗶𝝾𝗻. He'd traced one unknown app back to a whole pattern, dozens of tools staff had signed up for on their own, plenty with AI baked in that nobody had clocked. People stopped debating whether the risk was overblown and started asking how to clean up their own corner of it. Give people something concrete to look at, and accountability tends to follow. That, to me, is what enabling AI safely looks like in practice. Check out the full blog and how Reco gave them one view of everything: https://lnkd.in/eDzavFy6
-
Ofer Klein shared thisClaude Mythos changes one thing for SaaS teams: The time between a weakness existing and that weakness being exploited. The worry is not the model itself, but what a capability like this does once it points at applications and identities that were exposed all along. A weakness that looked isolated last year can now be chained to another in seconds, so an identity exposure and an over-permissioned integration join into one path instead of two separate tickets. Your team used to have days to investigate and remediate a new risk. As AI speeds up the analysis and the exploit, that window shrinks toward minutes. A misconfigured app is an easier target than hardened infrastructure, and AI spots those openings across your entire ecosystem at once. Non-human identities finish the picture. A service account with broad access to business-critical systems, and no governance around it, becomes a direct line to sensitive data. The strategy your team runs still works. What has to change is the clock. Close the exposure you own before Claude Mythos closes the distance.
-
Ofer Klein shared thisAdding a faster scanner to your stack right now would make your vulnerability problem worse. Discovery was never the bottleneck. Teams already process findings from every scanner and monitoring platform they run, and AI-scale discovery multiplies that volume past what any team can investigate. Without automated prioritization, the handful of findings that deserve immediate action sink under thousands that don't. Scoring makes the pile harder to sort, not easier. Traditional tools evaluate each vulnerability in isolation, on severity and technical traits, and a critical score says little about business risk, which is why federal guidance has moved to risk-based prioritization. Whether a finding gets touched first depends on context the scanner never sees, starting with who can reach the flaw and what data sits behind it. Then everything downstream of discovery runs on people. A finding still has to be validated and assigned, then fixed and verified, and that work scales at the pace of the humans doing it while the discovery side accelerates without them. And almost nothing correlates across the stack. The average enterprise runs dozens of disconnected security tools, so no single picture of exposure exists, and someone builds one in a spreadsheet, or nobody does. Speed at the front of the pipeline is solved. The rest of the pipeline is the work now.
Ofer Klein commented on a post
3d
Good luck Gur Talpaz!
-
Ofer Klein liked thisThe security model has to change because the environment has changed. Apps, APIs, and agents are being introduced faster than security teams can manually discover, assess, and monitor them — and the risk lives across the identities, permissions, and connections between them. Point-in-time visibility is no longer enough. Security needs to move at the speed of the ecosystem it’s protecting. That’s where AI changes the game — and that’s exactly Why Reco!!! 🚀
-
Ofer Klein liked thisOfer Klein liked thisI'm excited to share something we've been building at Tessl: Tessl Academy. Most of us are learning agentic development by experimenting with context files, skills, review loops and evals, and gradually figuring out what works. That's a good way to learn. But there's also real value in putting some structure around it. Tessl Academy is a hands-on curriculum for engineers who are already using coding agents and want to get better at engineering the context around them. You can work through it on the web, or (more fun, in my view) install it and have Claude Code, Cursor, Codex or Tessl Agent guide you through the lessons directly. You'll get hands-on with things like: - Writing skills your agent can actually use - Evaluating and optimizing your context to improve agent behaviour - Building agentic code review loops around your own conventions Some of it will reinforce what you already do. Some, I hope, will nudge you toward new practices you can develop deliberately and reliably. We're genuinely keen to hear how it lands. Try it, tell us what worked, what didn't, and what you'd like us to teach next. https://tessl.io/academy/
-
Ofer Klein liked thisOfer Klein liked thisWe just delivered one of our best quarters ever at Salesforce, outperforming across every key metric. 📈 $11.3B Revenue. +11% 💰 $33.5B cRPO growth. +14% 🚀 $5.90 Non-GAAP EPS. +103% 💵 $15.2B TTM Free Cash Flow 🤖 $3.9B Agentforce + Data 360 ARR ⚡ 7.0B AWUs delivered all-time 📊 Raising FY27 guidance to $46.1–$46.4B 🔥 3.2B AWUs in Q2 (+97% Q/Q) 🔗 6x growth in MCP calls 🤝 Customer attrition near all-time low Despite all the talk of a SaaSpocalypse, we’re seeing just the opposite. Net new AOV growth is the strongest it’s been in four years; attrition is near its lowest level ever; seats are growing, customers are upgrading to higher tier offerings; and agentic use of our platform is surging. AI isn’t replacing Salesforce. It’s unlocking more value across our platform. With this momentum, and Contentful and Fin joining Salesforce in Q3, we’ve raised our FY27 revenue guidance by $300 million in constant currency. We also announced Claudeforce, an exciting expansion of our partnership with Anthropic. Claudeforce combines Claude's extraordinary intelligence and reasoning with the trusted data, workflows, business logic, and governance businesses rely on. Powered by AIforce, our enterprise AI harness, Claudeforce brings the two together in a dynamic interface that makes Salesforce available wherever you work. This is the future of enterprise software. What Claudeforce changes today: • Instant Grounded Intelligence — ask complex enterprise questions and get real-time, trusted answers. • App & Agent Builder — deploy custom workflows, agents, and secure apps on the fly. • Action-Oriented — trigger live enterprise actions and get work done, not just summaries. • Ironclad Governance — Zero Data Retention, full trust boundary, Salesforce-certified by default. Probabilistic models alone can’t run a company. Deterministic systems alone can’t reason. Claudeforce fuses Claude’s reasoning with Salesforce’s trusted data and controls. The AI is the interface. This is how every business will run. See it at Dreamforce.
-
Ofer Klein liked thisOfer Klein liked thisis the SaaSpocalypse truly over? Marc Benioff certainly thinks so, and Salesforce shared some some numbers to back it up: → 3.2B agentic work units in Q2, up 97% q/q → 6x growth in MCP usage → top AI companies expanded their Salesforce spend 435% YoY 🤯 → seats are still growing enterprises still need systems of record for permissions, business rules, audit trails and governance EVEN in a headless world Read more in What's 🔥 in AI/Infra/VC #513 (full post in comments)
-
Ofer Klein liked thisOfer Klein liked thisToday, for many reasons, is a surreal moment for me. After years of discussing the art of the possible, I’m excited to unveil Cognition Venture Partners, a new early-stage fund I’m co-founding with Tayler Sipperly and Elia Zaitsev. There are so many exciting things about Cognition and its future. The biggest, however, is the ability to build with two individuals I hold in the highest possible regard. First of all, the chance to build something with Tayler as a true equal partner is the culmination of more than six years working together in the trenches across nearly every aspect of early-stage cyber, from ideation and origination to M&A execution. Those of you who’ve had the privilege of working with Tayler know there is no one more deserving of the title “partner” than him. He is, simply put, a force of nature who always stands by his word. He has earned his reputation with a no-shortcuts approach and his compass always points to founders as his North Star. I have always seen him as my equal (if not my intellectual superior) - this is just making it official. The fact that I’m partnering with Tayler is likely of little surprise to most of you, but Elia is a name I imagine most were not expecting. My relationship with Elia goes back nearly a decade to when I was an equity analyst convinced that CrowdStrike was going to change the world (spoiler alert: it did and continues to do so). Beyond being one of the sharpest minds I’ve had the privilege of working with, Elia is also amongst the most charismatic and endearing. He has a unique way of translating the hyper-complex into something easily digestible, and the industry still talks about his product videos (though Elia before the beard is a strange sight). I don’t believe there is anyone more valuable to early-stage companies than Elia, and the chance to partner together to build something in our own vision is an absolute dream come true. Our thesis for Cognition is simple - a no shortcuts model focused on hands-on partnership to help build the platforms of tomorrow. We’re building quickly with much, much more to come. If you have a chance, please read my partners Elia Zaitsev and Tayler Sipperly’s thoughtful posts. They do a far better job than I ever could explaining how we got here and where we’re going.
-
Ofer Klein liked thisOfer Klein liked this🚀 Big week for us. Incredible earnings and an incredible co-innovation with our friends at Anthropic. This kind of momentum doesn't happen in a vacuum — it's a direct reflection of the work our team has put in, and the strength of the partnerships we're building along the way. https://lnkd.in/gFDKX56ySalesforce and Anthropic Announce Claudeforce: The #1 AI Meets the #1 AI CRMSalesforce and Anthropic Announce Claudeforce: The #1 AI Meets the #1 AI CRM
-
Ofer Klein liked thisOfer Klein liked thisComing out of stealth mode… I'm excited to share that I'm joining Verizon as the incoming CISO. The opportunity to join Verizon at this moment was incredibly compelling. Verizon sits at the heart of critical infrastructure, connecting people, businesses and governments around the world, while undergoing a significant transformation focused on better serving customers and embracing an AI-first future. AI is changing how we build and operate technology, but also the speed and sophistication of the threats we face. I'm excited to continue working at that intersection: enabling responsible AI adoption, strengthening resilience against AI-enabled threats, and protecting infrastructure millions depend on every day. There are natural parallels between Cisco and Verizon. Both operate at enormous scale, are deeply connected to critical infrastructure, and have the opportunity to use their visibility and reach to improve security beyond their own four walls. I'm excited to carry many of the lessons from Cisco forward across an even broader ecosystem. I'm incredibly grateful to Nasrin Rezai for the security program and team she has built, and especially for her support and partnership during the transition. I look forward to building on that foundation and wish her the absolute best in her next chapter. Leaving Cisco was not an easy decision. I'm incredibly proud of what our team accomplished, but what I'll remember most isn't a technology, program or metric. It's the people and culture we built. We created an organization where people challenged conventional thinking, were transparent when things weren't working, broke down barriers, and came together when it mattered most. We navigated some incredibly difficult challenges and still managed to have a lot of fun along the way. Which brings me to the wolf. Among my going-away gag gifts, the team decided they needed something more memorable. So they "borrowed" one of the fake wolves positioned around the Cisco RTP campus to deter geese, wrapped it up, and presented it to me. So, briefly, my Cisco legacy included possession of stolen goose-deterrence technology. For the record, the wolf was returned safely to active duty. 😊 As ridiculous as it sounds, the wolf represents something I'll miss tremendously: a team that could tackle incredibly serious problems without taking ourselves too seriously. That combination of mission, trust, camaraderie and humor is difficult to manufacture. To my Cisco teammates, thank you for your trust, candor, friendship and partnership. You made me a better leader, and I'll always be rooting for you. I also owe a debt of gratitude to Anthony Grieco for his partnership and leadership during my time. It was fun and look forward to the continued partnership! And to my new Verizon teammates, I'm incredibly excited for what's ahead and grateful for the opportunity to join you. Time for the next chapter. #onward
Experience & Education
Other similar profiles
Explore more posts
-
HexaF Innovations
January was a strong execution month for HexaF Innovations. As an early-stage deep-tech company operating in the federal and aerospace ecosystem, we prioritize steady validation, operational discipline, and alignment with mission needs over headline growth. January milestones included: Funding & Institutional Validation • Awarded the City of Champaign New Business Grant • Secured a Tier-1 university grant Infrastructure & Scale Readiness • Accepted into the University of Illinois Research Park • Selected for the NVIDIA Inception Program Capability & Governance • Welcomed a Medical Advisor, Dr. Bilal Gondal, supporting space medicine and human performance efforts. Federal & Mission Engagement • Participated in the NASA cFS Symposium • Participated in the Agile Mission Assurance Workshop Grateful for the support across local (City of Champaign, Champaign County EDC, Illinois APEX Accelerator, Champaign Public Library, #UIUC FAST Center), academic, and national innovation ecosystems. We remain focused on building reliable, mission-relevant AI capabilities through long-term, trusted partnerships. #DeepTech #AI #MissionResilience #SpaceWeather #SpaceMedicine #HumanPerformance #SpaceTech #UIUC #UIRP #NVIDIAInception
-
Karan Bhatia
Array Labs Raises $20 Million to Scale Radar Manufacturing, Prepare for Launch. Array Labs, designing, building, and operating the world’s most advanced radar satellites, led by Andrew Peterson, Jose Isaac Robledo, and others, has announced a $20M Series A financing led by Catapult Ventures, with participation from Washington Harbour Partners LP, KOMPAS VC, and other new and existing investors, including Y Combinator, Maiora Capital, Animal Capital, Aera VC, Cultivation Capital, and Clearance Ventures, bringing Array’s total funding to $35M since going through Y Combinator. Read More At: https://lnkd.in/gnFEfyuH X: https://lnkd.in/g9WySuyu Source: https://lnkd.in/gga5NcZ3
-
Alisa Coffey
When the mission can’t fail, memory matters! Frontgrade’s new SpaceStor 3U SpaceVPX Solid-State MMU is purpose-built for harsh space environments. It delivers rugged, rad-tolerant memory in an open systems format that keeps critical data secure under extreme conditions. Frontgrade Technologies #space #storage #MMU #goldendome
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top contentOthers named Ofer Klein
22 others named Ofer Klein are on LinkedIn
See others named Ofer Klein