The First OWASP Top 10 Backed by Real Incident Data

Rock Lambros, Co-Lead of the 2026 OWASP Top 10 for LLM Applications, on why prompt injection would have fallen out of the top 10 based on incident data alone, and why the community kept it at number one anyway.In this episode I sit down with Rock Lambros, longtime security leader, former CISO, core team member of the OWASP GenAI Security Project, and Co-Lead of the 2026 OWASP Top 10 for LLM Applications, to dig into the new list and the harder questions underneath it. We get into the letter from the project leads that opens the document, why the LLM and Agentic lists are converging, what a year of messy incident data actually told the working group, and why Rock argues that agency is not authorization.In this episode:Why the LLM Top 10 and the Agentic Top 10 are merging in practiceThe opening letter, and why the industry was slow to admit the model was never the thing to secureWho benefits from a model-centric framing of AI securityPrompt injection ranked number one by practitioners and out of the top 10 by the incident dataWhy incidents get reported by outcome rather than by initial vectorMisinformation, the widest gap between the vote and the dataHidden Context Exposure, soft guardrails, and why instructions and data share one context windowContext rot and why context window management matters for agents doing critical workUsing AI to govern AI, and the dual-layer approach in the Agentic Control StandardAgency versus authorization, and why OAuth is what we have rather than the answerHow to get involved in the OWASP GenAI Security ProjectChapters: 0:00 Intro and Rock's background 1:17 What the agentic work changed about the LLM Top 10 3:59 The opening letter and "It Was Never the Model" 8:48 Incident data and the prompt injection ranking 11:23 Misinformation and the limits of the data 14:24 Hidden Context Exposure and soft guardrails 17:18 The context window and context rot 19:57 Using AI to govern AI 22:49 Excessive agency, and agency versus authorization 27:09 How to get involved with OWASPConnect with Rock:LinkedIn: https://www.linkedin.com/in/rocklambrosOWASP GenAI Security Project: https://genai.owasp.orgGet involved: https://genai.owasp.org/contributingOWASP Top 10 for LLM Applications 2026: https://genai.owasp.org/resource/owasp-genai-llm-top-10-2026/Read my piece on the new list, It Was Never the Model: https://www.resilientcyber.io/p/it-was-never-the-modelResilient Cyber: https://www.resilientcyber.io

Ondrej Vlcek, CEO of AISLE and former CEO of Avast, on why AI vulnerability discovery is not as commoditized as the industry thinks, and why remediation is still the real bottleneck.In this episode I sit down with Ondrej Vlcek, Founder and CEO at AISLE. Ondrej spent roughly 30 years in cybersecurity, joining Avast as employee number six or seven doing kernel-mode driver work on Windows 95, eventually becoming CTO and then CEO, taking the company public and selling it to NortonLifeLock in a nearly $9 billion transaction. He co-founded AISLE in the fall of 2024 to close the loop from discovery through triage, remediation, and verification. His team has now disclosed 350 plus CVEs across projects like OpenSSL and curl.We get into why the moat sits in the system and not the model, why the gray market price of vulnerabilities has not collapsed even as models get cheaper, and what it actually takes to ship a patch a maintainer will accept.In this episode:- Going from Avast intern to CEO, and why vulnerability management was the next problem- The jagged frontier, and why bigger models do not always mean better results- Which classes of bugs got cheap to find and which are still genuinely hard- Why vulnerability prices have not collapsed despite all the model progress- Building a model-agnostic system with bespoke benchmarks for model selection- Sovereign AI, on-prem and air-gapped deployment, and why findings are the real crown jewels- Triage, reachability, and why most findings are not actually exploitable- Patch verification, regression risk, and mitigations for embedded systems that cannot be patched- How AISLE earned trust from curl after Daniel Stenberg killed the bug bounty- Whether a CVE count is a vanity metric- Build versus buy as model capability keeps getting cheaper- What breaks first in the CVE and open source maintainer ecosystem- What AppSec leaders should change next quarterChapters0:00 Intro0:24 From Avast employee number six to a $9 billion exit3:26 Why vulnerability management, and why now5:15 The jagged frontier and what bigger models miss10:36 The economics of finding bugs, and why prices have not collapsed12:11 Building a model-agnostic system with real benchmarks14:30 Sovereign AI, air-gapped deployment, and who sees your findings19:42 Triage, reachability, and why remediation is the bottleneck24:48 Patches that break things, and systems you cannot redeploy25:39 curl, Daniel Stenberg, and death by a thousand slops29:35 Is a CVE count a vanity metric?31:34 Build versus buy when capability keeps getting cheaper34:41 What breaks first in the next 18 months39:46 What AppSec leaders should do next quarter41:13 ClosingOndrej Vlcek on LinkedInAISLEAISLE research and blogResilient Cyber SubstackSubscribe for more conversations with security practitioners and leaders.

Lovable CISO Igor Andriushchenko on soft guardrails vs. hard boundaries, securing vibe coding for non-developers, and building a security program at a 10x company.I sit down with Igor Andriushchenko, Head of Security and CISO at Lovable, the AI development platform behind one of the fastest growth stories in the space. Igor joined as the first security hire when the company was around 40 people. A year later he is running a 20+ person team covering product security, GRC, IT, and platform safety for a company with 400 laptops in MDM and no sign of slowing down.We get into what it actually takes to secure AI-native development, both inside a hypergrowth startup and on a platform where most of the people shipping software are not developers and definitely not security practitioners.In this episode:Building a security program for the company you will be in 12 months instead of the one you are in todaySoft guardrails versus hard guardrails, and how to decide which one a problem deservesWhy hard blocks push AI-assisted workflows into the shadowsRooting guardrail decisions in business goals, risks, and threats rather than tool defaultsDemocratized development without democratized security, and what a platform owes the 99%Lovable's auto-fix toggle, per-app threat models, and the goal of an app with no security tab at allWhether models will ever produce secure code by default, and why defense in depth still carries the loadGoverning the reality that every employee vibe coding an app looks a lot like a new vendorGRC engineering as the way to measure control efficiency layer by layer against AI-powered attackersCRA, NIS2, and the EU AI Act landing on citizen developers who never thought of themselves as software manufacturersChapters: 0:00 Intro 0:23 Igor's background from DevOps to CISO 3:54 Scaling security at a 10x company 6:07 Reinventing the team when growth breaks it 08:26 Soft guardrails versus hard blocks 14:05 Tying guardrails to business risk 17:32 Democratized development, undemocratized security 18:52 Shared responsibility on an AI dev platform 21:16 Auto-fix, per-app threat models, and no security tab 25:21 Will models produce secure code by default? 29:56 Every employee vibe coding is a new vendor 30:57 Enterprise controls, publishing gates, and PII scanning 36:39 AI-powered attackers and why good enough changed 40:43 GRC engineering and measuring control efficiency 46:19 CRA, NIS2, and the citizen developer 52:41 Trust centers for builder apps 54:08 Closing thoughts on the vibe coding communityGuest links: Igor on LinkedIn: https://www.linkedin.com/in/igor-andriushchenko Lovable: https://lovable.devResilient Cyber: Newsletter and episode archive: https://www.resilientcyber.io Subscribe for more conversations with security practitioners and leaders.

CISOs have a stack of tools but no system built to run the security program itself. Mike Armistead wants to fix that.In this episode I sit down with Mike Armistead, co-founder and CEO of Pulse Security AI and a longtime security founder behind Fortify and Respond Software. We dig into why the security leader has never had a system of truth the way the CFO has an ERP and the CRO has a CRM, and how an agentic layer on top of the existing tools can finally close that gap. Mike is measured about where AI gets to decide and where the human stays in the seat, and he shares what surprised him most from research with more than 80 senior practitioners and corporate directors.In this episode:- Why two exits later Mike came back to build a third company around the AI wave- The silos that left CISOs with an acronym soup of tools and no way to run the program- What a system of truth for the CISO actually means and how it layers on top of existing structured and unstructured data- Where agents do the heavy lifting on regulatory monitoring, vendor intelligence, and status reporting- Governing the guardrails, not the keystrokes, and why closing the loop still involves people- What corporate directors actually want to hear in the 15 to 20 minutes a CISO gets each quarter- The findings that stood out, including that 55% of boards have never defined the cyber risk they are willing to accept, and only 12.5% of CISOs are very confident the board leaves with a true picture of the risk- Institutionalizing the tribal knowledge every security program runs onChapters:0:00 Intro0:18 Mike's background and two prior exits1:08 Why the AI wave pulled him back2:21 Why the CISO has no system to run the program4:09 Starting at the program level, not the SOC or AppSec5:28 What a system of truth for the CISO means8:19 Speaking the language of the business9:09 Where AI does the heavy lifting on a typical Tuesday11:57 Govern the guardrails, not the keystrokes15:44 Bringing deputies into the conversation16:46 What the research with senior practitioners found20:37 Boards, risk tolerance, and the reporting gap24:57 AI as a double-edged sword for security leaders25:35 Joanna Burkey and institutionalizing tribal knowledge27:31 A year from now for the security leaderGuest links:Mike Armistead on LinkedInPulse Security on AIMore Resilient Cyber:Substack: https://www.resilientcyber.ioSubscribe for more conversations with security practitioners and leaders.

Alex Pinto, who leads Verizon's DBIR team, joins me to break down the new Breach Impact Study and what data breaches actually cost organizations.For years the industry has argued past itself on breach costs. One camp says the market doesn't care, the other says a single breach ends your business. Alex and his team finally got their hands on roughly 70,000 cyber insurance claims through CyberAcuView, and the Breach Impact Study puts real numbers behind the question. In this conversation we dig into what the data shows, where it stops, and how a security leader should actually use it.Alex Pinto runs the Data Breach Investigations Report team at Verizon Business and has been building the report for close to a decade. The Breach Impact Study is the team's first focused spin-off from the DBIR.In this episode:- How the Breach Impact Study came together and why the DBIR team finally got cyber insurance claims data- Why the study measures insurable loss as a floor, not a ceiling, of real economic impact- The case for reporting medians over averages, and why the team refuses to publish the average- Business interruption versus contingent business interruption, and why downtime moves the needle- Whether an $83,000 median breach impact sends executives the wrong message- The SMB paradox, where the smallest companies take the hardest proportional hit- What the claims data does and does not show about AI on offense and defense- Third-party risk, coverage sub-limits, and the single biggest takeaway for security leadersChapters0:00 Intro0:24 Meet Alex Pinto and the DBIR team2:51 Launching the Breach Impact Study3:26 Getting cyber insurance claims data7:32 Why insurable loss is a floor, not a ceiling11:14 Medians over averages, and why the average is meaningless15:13 Business interruption vs contingent business interruption19:49 Does an $83K median send the wrong message?22:44 The SMB paradox and the cybersecurity poverty line26:05 Where AI shows up, offense vs defense34:48 The CVE explosion and marketing hype36:59 Third-party risk and coverage limits41:34 Wrap-upGuest linksAlex Pinto on LinkedIn: https://www.linkedin.com/in/alexcpsec/Alex Pinto on X: https://x.com/alexcpsecVerizon DBIR and Breach Impact Study: https://www.verizon.com/business/resources/reports/dbir/More from Resilient CyberSubstack: https://www.resilientcyber.ioSubscribe for more conversations with security practitioners and leaders.#cyberrisk #databreach #cyberinsurance #ransomware #aisecurity #dbir

What happens to security investing when vulnerability discovery becomes continuous and exploitation windows shrink from weeks to hours? I sit down with Chenxi Wang of Rain Capital to dig into it.Chenxi is the Founder and Managing General Partner at Rain Capital, a venture fund focused on early-stage cybersecurity companies. She's been a Carnegie Mellon professor, a Forrester VP, and a strategy leader at Intel Security and Twistlock, and her portfolio includes companies like Claroty, ProjectDiscovery, Ox Security, runZero, and Straiker. She closes out my July run of conversations with security investors.In this episode:The AI Exploit Age and why vulnerability discovery is becoming continuousGuardian Agents and the case that it takes an AI to govern an AISeparating AI agent identity from traditional machine identityThe signals that predict enterprise adoption for early-stage security startupsThe barbell funding market and the squeeze on Series B and CWhat security leaders should do differently over the next twelve monthsConnect with Chenxi: LinkedIn: https://www.linkedin.com/in/chenxiwang88/ Rain Capital: https://raincap.vc/ Rain Capital Insights: https://raincapital.substack.comSubscribe to Resilient Cyber for more conversations with security practitioners and leaders: https://www.resilientcyber.io

Bugcrowd founder Casey Ellis joins me to dig into what AI is actually doing to bug bounties, vulnerability discovery, and open source security. We get into his "slopdemic" framing, the curl bug bounty saga, VDP readiness, the pentest market correction, and where security research policy heads next.Casey Ellis is the founder of Bugcrowd, co-founder of disclose.io, and a board member of the Security Research Legal Defense Fund. These days he advises and invests through Tall Poppy Group and works at the intersection of security, AI, and policy. His argument is that the vulnpocalypse was already here, and AI has made the cost of both finding and reporting vulnerabilities collapse at the same time.In this episode:Casey's path from building Bugcrowd to advising, investing, and policy workWhy more practitioners need to get involved in policy, and why law is just codeThe slopdemic vs. the vulnpocalypse, and what actually changed in submissionsAI lowering the bar for a broader, less predictable pool of threat actorsDaniel Stenberg, curl, and maintainers below the security poverty lineThe lightning rod vs. rockets distinction between VDPs and bug bountiesThe pentest market correction underway from AI pricing pressureCollapsing OODA loops, hack-back, CFAA reform, SRLDF, and disclose.ioChapters:0:00 Intro and Casey's background 2:56 Why practitioners belong in policy 6:22 The slopdemic vs. the vulnpocalypse 9:40 AI lowering the bar for threat actors 11:47 Open source, curl, and the security poverty line 15:37 VDP vs. bug bounty readiness 19:20 The pentest market correction 24:20 What breaks first in vulnerability management 27:20 Hack-back and non-cooperative defense 28:43 A near-term playbook for security leaders 31:40 CFAA, SRLDF, and disclose.ioConnect with Casey: LinkedIn: https://www.linkedin.com/in/caseyjohnellis Blog: https://cje.io disclose.io: https://disclose.io Bugcrowd: https://www.bugcrowd.comResilient Cyber: https://www.resilientcyber.io Subscribe for more conversations with security practitioners and leaders.

Jon Sakoda of Decibel joins me to break down AI's impact on cybersecurity startups, venture funding, and why endpoint is the Super Bowl of cyber.Jon is the Founding Partner at Decibel, an early-stage firm backing technical founders in security and infrastructure. He started his career founding IMlogic, an IM security company acquired by Symantec, then spent over a decade at NEA working with companies like Cloudflare, MongoDB, and HackerOne before launching Decibel. We got into why he thinks AI is only magical if you have a magic power, why Decibel led a $100M seed into Ent, and where the firm is placing its next bets.In this episode:Why Decibel operates like the Navy SEALs next to the big platform fundsThe founder community model and finding the early believers among CISOsWhat separates the founders who finish now that AI lets everyone startEnt's $100M seed and the self-driving moment for endpoint securityTelling genuinely AI-native companies apart from AI washingAI eating venture capital and why cyber's best years are aheadOpen models, frontier labs, and why the cat is out of the bagThe agentic SOC, Dropzone AI, and driver assistance vs. self-drivingStartup consolidation cycles and being an N of oneHow buyers and job seekers should evaluate early-stage vendorsDecibel's next bets, from novel AI models to resilience and cyber insuranceChapters:0:00 Intro 0:32 Jon's background and founding Decibel 2:25 Big platform funds vs. specialized firms 3:56 Founders helping founders and early believers 6:22 Scaling beyond the early adopters 7:40 Who finishes the marathon in the AI era 10:19 Founders from outside cyber 12:21 Ent's $100M seed and the endpoint bet 14:53 AI-native vs. AI washing 17:04 AI is eating venture capital 18:55 Open models vs. frontier labs 22:41 The agentic SOC and Dropzone AI 26:03 Consolidation and the startup cycle 29:22 How buyers should evaluate young vendors 31:43 Decibel's next bets and cyber resilience 34:11 Game Day at Black HatConnect with Jon: LinkedIn: https://www.linkedin.com/in/jonsakoda/ Decibel: https://www.decibel.vcSubscribe for more conversations with security practitioners and leaders, and find my writing at https://www.resilientcyber.io

In this episode I sit down with Ed Sim, founder and managing partner of Boldstart Ventures, to dig into where AI security, agentic infrastructure, and the venture market are actually heading.Ed has been an inception-stage investor for nearly 30 years and has run Boldstart since 2010, backing hardcore technology companies across AI infrastructure, cybersecurity, and physical AI. He was the first investor in Protect AI, which sold to Palo Alto Networks in a reported ~$700M exit roughly a year before ChatGPT launched. He is also early in companies like Keycard, Surf AI, and June. About a third of Boldstart's investments are in cyber, so Ed sees this market from the founder and investor side in a way most security conversations do not.We get into why the era of building raw intelligence is giving way to an era of controlling it, what that means for on-prem models and private evals, and why Ed thinks nearly everything in security is going to get rebuilt from scratch.In this episode:- Why a day-one partnership looks different now that anyone can vibe code an MVP- The Protect AI acquisition and what the first exit in AI security signaled to the market- Competing as an inception fund against mega-funds writing giant seed rounds- What founders should actually look for in a venture partner beyond the check- The shift from building intelligence to controlling it, including routing, post-training, and on-prem deployment- Why enterprise data, workflows, and private evals are becoming the crown jewels- Vulnerability chaining, attack path reasoning, and how tools like Mythos are reshaping the security budget conversation- Agentic identity and why Keycard treats agents as short-lived problem solvers rather than digital twins- The Surf AI thesis on automated security hygiene and tying every asset back to an owner- The real bottleneck slowing agent adoption in the enterpriseChapters:0:00 Intro0:35 Ed's background and inception investing1:57 Day-one partnerships in the vibe-coding era3:53 The Protect AI exit to Palo Alto6:14 Competing as an inception fund against mega-funds9:17 What founders should look for in a VC partner11:48 From building intelligence to controlling it15:52 Boldstart's domain-specific model portfolio16:16 Private evals, context, and memory as crown jewels17:18 Mythos, vulnerability chaining, and attack path reasoning20:59 How much access should you give the model22:07 On-prem context and the autonomous workforce24:49 Agentic identity and Keycard28:11 Building brand and community with Insecure Agents31:30 The Surf AI thesis and automated security hygiene34:13 The real bottleneck to agent adoption37:09 The easy button, Palantir, and a multi-model world38:24 Two types of people in this new eraConnect with Ed:LinkedIn: https://www.linkedin.com/in/edsim/Boldstart Ventures: https://boldstart.vcEd's newsletter, What's Hot in Enterprise IT/VC: https://www.whatshotit.vcMore from Resilient Cyber:Substack: https://www.resilientcyber.ioSubscribe for more conversations with security practitioners and leaders.#aisecurity #agenticai #cybersecurity #venturecapital #appsec

Does restricting frontier AI in the name of safety actually make us less secure? Joshua Saxe joins me to make the case that it does, and that AI cybersecurity will be won through defender adoption, not restriction.Josh has spent 15 years at the intersection of AI and security. He built and ran the machine learning program at Sophos, then led security for Llama at Meta, covering security post training, evals, agent guardrails, and prompt injection prevention. He recently left to co-found a startup reimagining vulnerability and exposure management agentically. He also writes one of the most cited blogs on AI and cyber policy.In this episode:- Why restricting frontier model access harms defenders more than attackers- How monitored closed models put threat actors at a structural disadvantage- The jagged frontier, and why attackers don't need frontier models for most of their tradecraft- The national security and supply chain risks of pushing the world onto Chinese open weights models- Why exploits don't cause cyberattacks, and which attacker constituencies AI actually unblocks- The dual use ceiling on guardrails and classifiers- Where defenders should be adopting AI right now, from access management to SOC automation- Using agents to burn down the mountain of security technical debtChapters:0:00 Intro0:42 Josh's background, from blackhat teen to Llama security lead3:07 The case for diffusion over restriction6:14 Why restriction hurts defenders more than attackers10:19 The jagged frontier and what attackers actually use models for12:49 National security and the supply chain risk of Chinese open weights16:08 Exploits don't cause cyberattacks20:20 Where defenders should adopt AI right now24:20 Guardrails, classifiers, and the dual use problem27:34 Reimagining vulnerability management with agents32:17 The structural advantage defenders hold35:15 Policy wishes and the attacker's Claude Code momentFollow Josh:LinkedIn: https://www.linkedin.com/in/joshua-saxe-01845a1Substack: https://joshuasaxe181906.substack.comFollow Resilient Cyber:Substack: https://www.resilientcyber.ioSubscribe for more conversations with security practitioners and leaders.#aisecurity #cybersecurity #vulnerabilitymanagement #aipolicy #opensourceai