LuminaTalks Podcast

LuminaTalks Podcast

by Kevin De Pauw
Technology
Season 2
How an AI contract can cost your business a $2M deal | Corporate Lawyer Ruben Schoenmaekers explains
What happens when founders start using AI to manage legal decisions and something goes wrong? In Episode 9 of LuminaTalks, I sat down with startup lawyer Ruben Schoonmakers. We discussed how founders are actually using AI right now, what's working, and where things quietly go sideways. From early-stage founders copy-pasting contracts and patching them with AI tools, to scaleups trying to close million-dollar deals without the right legal infrastructure in place. Ruben's take? AI can flag the risk. It can't own it. If you're building a company, investing in one, or just trying to understand how AI is reshaping business decisions, this one's worth your full attention. Ruben shares his honest predictions on AI, legal careers, and the future of startup strategy. 🔔 Subscribe to LuminaTalks for real conversations on AI, startups, business strategy, and the future. Get in touch: LinkedIn: https://www.linkedin.com/in/depauw/ Instagram: https://www.instagram.com/luminatalks/ Summ.link: https://summ.link/
S2 · E9
Mar 27, 2026
1:20:14
War is the Best Driver for Technological Advancements | Raf Lefever
Is AGI the new Manhattan Project? The nuclear race was won with secrecy and science. But the AI race is going to be won with chips, energy, water, and cables. Everyone talks about AI like it's a software problem. Build a better model, win the race. But the people actually shaping the future know the real battle is happening in a completely different place, in the physical infrastructure that powers intelligence. Who controls the compute? Who controls the electricity? Who controls the data centres? That's what we get into this week. Raf and I cover a lot of ground on the global AI arms race, why Europe might actually have a shot at doing this differently, and a genuinely radical idea: what if your home could one day generate digital value the same way solar panels generate energy? We also tackle something that doesn't get nearly enough attention, fragility. A handful of damaged cables or a targeted attack on centralised infrastructure could take down entire economies. Is that a risk we should just accept? Or is there a smarter, more resilient way to build? If you're into AI, geopolitics, digital sovereignty, decentralisation, cloud infrastructure, or just where the global economy is actually heading — this one's worth your full attention. Listen now and let us know: Should infrastructure be centralised or distributed? Drop your take in the comments. Get in touch: LinkedIn: https://www.linkedin.com/in/depauw/ Instagram: https://www.instagram.com/luminatalks/ Summ.link: https://summ.link/
S2 · E8
Mar 13, 2026
1:11:29
Trump does not want a NUCLEAR BOMB to wipe Europe, He needs Azure & AWS | Raf Lefever Explains WHY
"We have to become independent from the United States. We are at war." In this very special episode I sat down with Raf Lefever and honestly, some of the stuff we talked about here kept me up at night thinking about it. So basically, we're diving into this whole mess about how Europe needs to break free from American tech companies, and Raf drops this bomb about how we're literally living under tech feudalism right now. Like, he compared cloud providers to drug dealers - which sounds crazy until you actually think about it and realize he's probably right. We're at war. Not with bullets, but with data, compute, and dependency. Right now a small group of tech CEOs in the United States effectively control the infrastructure the entire world runs on. Try pulling your data out of a major cloud provider sometime. You'll quickly realize you were never really a customer, you were a serf. That's not a metaphor I'm throwing around lightly. Raf breaks down exactly why what we're living through today maps frighteningly well onto medieval feudalism, where a handful of lords controlled the land and everyone else just worked it. The land today is compute. And most of Europe is farming someone else's. Trump, AI, cloud lock-in, digital sovereignty, we get into all of it. And yes, we talk about what Europe actually needs to do to stop sleepwalking into permanent dependency. This is the teaser. The full conversation is coming very soon and trust me, you don't want to miss it. Hit subscribe so you're the first to know when it drops and let me know in the comments whether you think Europe still has a real shot at this. #LuminaTalks #DigitalSovereignty #EuropeTech #TechnoFeudalism
S2 · E7
Feb 27, 2026
1:06:36
OWASP Expert reveals hidden vulnerabilities in AI agents | Vineeth Sai Narajala
AI is no longer just generating answers. It’s executing tasks, using tools, interacting with other agents, and operating with autonomy. In Episode 6 of LuminaTalks, AI Security Researcher Vineeth Sai Narajala (Cisco) and contributor to OWASP initiatives breaks down what it really takes to build secure, trustworthy agentic AI systems. This conversation goes into practical architecture decisions. What We Cover in This Episode: ✔️ What agentic AI actually means (beyond chatbots) ✔️ Why giving AI “agency” changes the security model ✔️ Prompt injection & multi-agent attack chains ✔️ Identity, responsibility & “Kevin 2.0” scenarios ✔️ Why the AI trust problem is more about identity than alignment ✔️ Human-in-the-loop vs sandboxing ✔️ Kill switches & blast radius reduction ✔️ Least privilege & just-in-time access ✔️ Why LLMs may not become a bigger security threat than web apps ✔️ The future of AI identity systems Vineeth shares insights from his work within OWASP. And we also explore how Agent Name Service (ANS) extends trust to AI agents by building on DNS and certificate-based identity systems. As millions of AI agents begin operating across ecosystems, identity, authentication, and verification become foundational. 🧠 10 Key Takeaways: → Use OWASPs Agentic Security Guide as Your Design Baseline → Stop Giving Agents Static Credentials → Implement a Memory Firewall for Your Agents → Always Separate LLMs from Tool Access Logic → Use OWASPs Agentic Threat Templates for Internal Red Teaming → Design for Detection: Don’t Just Block, Monitor → Adopt 'Andon Cord' Patterns to Kill Compromised Agent Sessions → Start Using Agent Fingerprints for Identity Control → Apply Least Privilege to Every Tool the Agent Can Access → Join OWASPs Agentic Security Initiative or Align With AIVSS/AISVS Early If you're building AI agents, designing autonomous systems, or leading platform security, this episode gives you practical insights to apply immediately. 🔔 Subscribe for deep conversations on AI security, agentic systems, governance, and building trustworthy AI infrastructure.
S2 · E6
Feb 13, 2026
49:26
Only 0.3% Founders will Reach €10M in Revenue | Jürgen Ingels Explains Why
Most founders believe startups fail because they run out of money. Experienced investors know that’s rarely the real reason. In this episode of LuminaTalks, I sat down with Jürgen Ingels, investor, entrepreneur, and founder of Supernova, one of Europe’s leading tech events. We discuss how companies actually scale, why focus matters more than funding, and how technology is quietly reshaping the way investors make decisions. Modern startups are being built with smaller teams, higher profitability, and radically different operating models. Yet Europe is falling behind in tech adoption, and how automation and AI are already influencing investment decisions long before a founder walks into a room. Why? This conversation goes into patterns, systems, and the uncomfortable truths founders need to understand if they want to build companies that last. In this episode, we talk about: ✔️ Why investors never fully trust your business plan ✔️ How experienced investors mentally rewrite startup projections ✔️ The rise of small, highly profitable teams and the “mushroom model” ✔️ Why hiring too early can slow your company down ✔️ How automation is changing back offices, finance, and operations ✔️ The growing role of AI in filtering and evaluating startups ✔️ Why Europe is structurally behind in tech — and what founders can learn from Asia ✔️ The difference between building products and building platforms ✔️ Why second- and third-generation founders matter for ecosystems 🧠 Key Takeaways from this episode: → Use a Clear Strategic Focus to Decide What You Ignore → Build Investor Relationships Long Before You Need Money → Pitch the Plan Investors Will Recalculate Anyway → Ask Investors to Remove Operational Load → Protect Founder Focus as a Core Responsibility → Learn Faster by Borrowing Experience → Use Events to Create Outcomes, Not Exposure → Build Platforms over Products → Observe How Your Team Disagrees Under Pressure → Deliberately Stress Test Founder Dynamics Early → Use Remote Work Strategically, Not Ideologically → Stay Lean Until Growth Forces Structure → Prioritize Focus Over Speed → Tell a Founder Story, Not a Feature List → Formalize Founder Agreements Before Success Whether you’re a founder, operator, or investor, this episode will challenge how you think about growth, scale, and decision-making in today’s tech landscape. 🔔 Subscribe to LuminaTalks for more conversations on technology, business, and the systems shaping the future.
S2 · E5
Feb 16, 2026
1:15:09
The BIGGEST Mistake You're Making with AI Right Now | Petar Tsankov
In this episode, I sit down with Petar Tsankov, AI safety researcher, engineer, and one of the leading voices on enterprise AI governance and compliance. We unpack why traditional, checkbox-driven AI compliance is breaking down, and why governance can no longer live in policies, audits, or PDFs. As AI systems scale faster than regulation, enterprises are being forced to rethink compliance as an engineering problem, not a legal one. This episode goes deep into how AI governance must shift from documentation to execution. We explore: ✔️ Why annual AI audits fail the moment models change ✔️ The biggest blind spot in enterprise AI governance and why most teams don’t see it ✔️ Why compliance must become continuous, automated, and embedded in pipelines ✔️ How abstract principles like fairness, transparency, and security break down at the technical level ✔️ Why AI bias is unavoidable and why measuring it early matters more than eliminating it ✔️ How regulation can enable innovation, not slow it down, when applied to high-risk use cases ✔️ What “from checkbox to code” actually means for AI teams in production 🧠 15 Key Takeaways from this episode: → AI governance began in academic research but has become a strategic business necessity, especially for large enterprises. → AI governance works only when it’s embedded into business strategy, not treated as a separate function. → AI forces compliance to shift from one-time checks to continuous, real-time evaluation. → Trust in AI depends on transparency. People need to understand how models are trained and how decisions are made. → The hardest part of AI governance is turning abstract ideas like fairness and security into measurable technical metrics. → Bias in AI is inevitable, but catching and measuring it early is what makes systems fairer. → AI governance must live inside development pipelines, not be added after systems are built. → The future of AI governance is code-based, with ethical principles built directly into systems, not paper audits. → Smart regulation can enable AI innovation if it focuses on risk without slowing scale. → The biggest AI governance blind spot is the gap between high-level principles and real technical implementation. → AI systems need continuous, automated validation because manual compliance can’t keep up with change. → Automation should make AI governance easier, reducing manual work without slowing innovation. → AI governance fails without strong data governance. If you can’t track your models and data, you can’t stay compliant. → High-risk AI systems demand stricter oversight and risk-based governance frameworks. → AI will reshape work by automating routine tasks and amplifying human intelligence, not replacing it. 🔔 Subscribe for more conversations on building AI that scales safely, responsibly, and in production. #AIGovernance #AICompliance #ResponsibleAI #EnterpriseAI #AIRegulation #EUAIAct #MachineLearning #AISafety #TechLeadership #FutureOfWork
S2 · E4
Feb 16, 2026
1:09:41
AI Safety Expert: When AI Understands You Better Than You Realise | Dr. Marta Bieńkiewicz
In this episode, we sit down with Dr. Marta Bieńkiewicz — neuroscientist, strategist, and one of the clearest voices in EU AI policy. We explore the ethics, tech and governance of AI agents, from VR neuro-rehab to delegation, identity, and whether agents can (or should) act on our behalf. We are entering the era of agentic AI: autonomous systems that don’t just execute instructions but represent you in decisions, negotiations, and even relationships. ✔️ Why “delegation” to AI often turns into abdication — and how that breaks systems of responsibility ✔️ How multi-agent AI ecosystems (like OpenAI’s hide-and-seek experiment) are evolving behaviors that humans didn’t design ✔️ Why trust is Europe’s strongest bet in the global AI race — and how to turn regulation into a strategic advantage ✔️ The dangers of emotional AI that simulate care without understanding, and how it can create ethical blindspots ✔️ How neuroscience is informing AI ethics — from brain signals to algorithmic decision-making 🧠 15 Big Takeaways From This Episode: AI alignment is not ethics. It ensures an AI does what humans intend, not what is most efficient → Alignment prevents systems from optimizing for the wrong outcomes. It is the core safety challenge for agentic AI. Human–AI interactions should include friction. Too much smoothness becomes addictive. → Perfect emotional mirroring makes AI feel more attentive than humans, increasing dependency and reducing user autonomy. Cyber sickness is still a real problem in VR/XR. Don’t drive immediately after using it. → VR affects balance, motion perception, and reaction time. Even short exposures can impair motor safety. VR/XR reduces real human gaze interaction. This impacts children’s social development. → Eye contact shapes emotional learning and bonding. Reduced real-world gaze time has long-term cognitive implications. Agents can misunderstand instructions. Always include clear boundaries and fallback rules. → Even “simple” automation can behave unexpectedly. Constraints prevent unintended actions and protect user interests. You must know what your agent knows. Information asymmetry is a serious risk. → If the system provider has more insight into your agent than you do, its loyalty may shift away from the user. Trust in AI equals predictability. If you cannot predict it, you cannot trust it. → A system that behaves inconsistently or unexpectedly cannot be relied upon in real decisions or delegation. Agents representing you can drift from your values if you do not observe their actions. → When agents make independent decisions, their internal model of “you” evolves. Over time, it stops reflecting your intent. Evolutionary AI and open-ended systems learn faster than humans can track. Oversight becomes impossible. → Such systems self-improve without human understanding, creating unpredictable and potentially unsafe emergent behavior. Multi-agent systems have three key failure modes: conflict, miscoordination, and collusion. → These patterns explain how multi-agent breakdowns occur and guide the design of safer interactions. Agents need normative infrastructure. Rules and shared constraints prevent chaotic behavior. → Just like human societies, agents require norms that shape what they can and cannot do. Geospatial multi-agent systems can improve environmental modelling, but still require careful governance. → Real-world predictive tasks (disasters, climate, resource use) benefit from multi-agent cooperation but introduce complexity and risk. Trust certification will become essential. → AI trust will not remain optional. Certification will become a competitive edge for products, teams, and platforms. Liability for agent mistakes requires IDs, logs, and transparency. → Auditable logs and digital identities are essential for determining who is responsible when agents cause harm. → Clear boundaries protect society. 🔔 Subscribe for more episodes on building AI that scales safely and responsibly.
S2 · E3
Feb 17, 2026
1:24:44
The AI Regulation Expert: Your Last Wake-Up Call on EU AI Act Rules | Kai Zenner
This one hits deep.Kai Zenner - one of the key minds shaping Europe’s AI policy - sat down with me for a real, no-BS conversation about how the AI Act might actually strengthen founders instead of slowing them down.We went everywhere: • What really counts as high-risk • The future of general-purpose models • Why “trust” might be Europe’s biggest export • How structure can become a scaling advantage for startupsThis isn’t politics. It’s about staying ahead while the rules are still being written. Watch, think, and tell me: Is Europe shaping the next wave of trustworthy AI… or just an overcomplicating path to building?
S2 · E2
Feb 16, 2026
1:32:01
Europe's Plan to Finally Unleash its Startups | 28th Regime with Greta Koch & Iwoana Biernat
EU-INC: What if Europe had one single system to start, fund, and scale startups? In this special episode, Greta Koch (European Parliament) and Iwona Anna Biernat (Project Europe) join host Kevin De Pauw to unpack the vision behind EU-INC! The “28th regime” designed to unify European company law and accelerate innovation. 💡 Topics include: • Why the EU-INC could change how startups are built across Europe • How the AI Act connects to innovation, governance, and trust • What Europe must do to stay competitive with the U.S. and China 👇 Do you support #EUINC? 💬 Let’s talk about how Europe can build the next generation of startups. 🇪🇺
S2 · E1
Feb 16, 2026
1:26:47