Trusted AI Agents: Architecting Identity and Granular Access for the Agentic Web
Nu AI-agenten steeds vaker autonoom over netwerken heen opereren, is het cruciaal om robuuste identiteits- en nauwkeurige toegangscontroles in te stellen om vertrouwen en verantwoorde interactie te bevorderen.
Deze LFDT Belgium Meetup onderzoekt de fundamentele uitdagingen van het veilig herkennen en delegeren van AI-agenten, en het mogelijk maken van gedetailleerde datatoegang en persistent agentgeheugen via innovatieve wallet-oplossingen.
Praktische info:
Leertrajecten
georganiseerd door:
Belangrijke onderzoeksvragen:
- Hoe kunnen agenten betrouwbaar worden geïdentificeerd en vertrouwd in uiteenlopende omgevingen?
- Welke delegatiemodellen zorgen ervoor dat agenten op een gepaste manier namens gebruikers handelen?
- Hoe kunnen gedetailleerde toegangsrechten het principe van minimale privileges waarborgen bij gegevensgebruik?
- Waar moet door agenten gegenereerde data veilig worden opgeslagen om interoperabiliteit en duurzaamheid te garanderen?
Programme
19:00 - 19:10 Welcome & introduction by Howest Cyber3Lab
19:10 - 19:40 Scaling the Agentic Web: New Challenges and Areas of Innovation
Andor Kesselman - Founder, DIF Labs
As multi-agent systems move from theory into deployment, one of the most urgent challenges is establishing agentic identity - how agents are recognized, authenticated, and trusted across diverse networks. This session will examine the complexities of delegation (how agents act on behalf of humans or other agents), the role of hardware attestations in anchoring trust, and emerging patterns of human–agent interaction. We will also discuss how agents differ fundamentally from broader AI systems and frameworks to think about them. Drawing on evolving work at the Decentralized Identity Foundation (DIF) and MIT’s Project NANDA, we will highlight emerging frameworks designed to scale the agentic web responsibly.
19:40 - 20:10 Agentic Wallets™: Enabling Fine-Grained and Persistent AI Data Access with Solid and MCP
Geoff Pirie - Director of Product, Inrupt
20:10 - 20:40 Panel Discussion
Shane Deconinck - Web3 Lead, Howest Cyber3Lab
Andor Kesselman - Founder, DIF Labs
Geoff Pirie - Director of Product, Inrupt
This panel will bring together Andor and Geoff to dive deeper into the challenges and opportunities in building trusted AI agents. Topics include the evolution of agentic identity standards, the practicalities of delegation models, and the importance of granular, consent-driven permissioning through Agentic Wallets.
Shane will facilitate a dynamic discussion on how these technologies intersect and complement each other to create a secure, interoperable, and human-centric agentic web. Attendees will have the opportunity to ask questions and explore real-world implications for developers, architects, and organizations aiming to adopt trusted AI agent frameworks.
20:40 - 20:45 Closing Remarks
~~~~~
Sluit je aan om te ontdekken hoe gedecentraliseerde identiteit, delegatie en gecontroleerde toegang kunnen helpen de uitdagingen op te lossen die de volgende generatie van vertrouwde AI-agenten vormgeven.
Lesgevers / sprekers
Andor Kesselman is a seasoned CTO and recognized leader in Decentralized Identity, with a deep focus on multi-agent systems and governance. He has been researching agent architectures since 2018, including early work in DecPOMDP that anticipated today’s advances. His startup, AgentOverlay, focuses on bridging multi-agent theory with years of work in decentralized identity. He helps lead the Bay Area chapter of Project NANDA and chairs the Technical Steering Committee at the Decentralized Identity Foundation (DIF), where he is also launching the Trusted AI Agents Working Group. His work centers on the emerging challenges of multi-agent governance-from identity and trust to policy enforcement-critical issues that must be solved to responsibly scale the agentic web.
