Online Trust in the Age of Agentic AI: Key Insights from Identity Week Europe 2025

This year’s Identity Week Europe 2025 gathered industry leaders, regulators, and innovators to discuss the opportunities and challenges in the identity space. Among the insightful sessions, Veriff’s CPTO, Hubert Behaghel, delivered a compelling keynote, “Online Trust in the Age of Agentic AI.” Follow along to learn about key takeaways from his talk.

As artificial intelligence evolves from passive tools into autonomous agents, the architecture of online trust faces a critical reckoning. While regulators move to govern developers and deployers, the wider digital ecosystem remains alarmingly unprepared for the scale and subtlety of AI-driven abuse.

In the keynote session, Hubert Behagel explored the emerging fault lines in our digital trust infrastructure, offering a structured, risk-based framework built around three foundational questions: Are you who you claim to be? Are you the rightful user of the account? Can you be trusted? As AI systems gain agency, we must now ask a fourth: Are you authorised?

Agentic AI: Amplifying the Trust Challenge

The rise of Agentic AI – systems capable of setting goals, making autonomous decisions, and taking effective action – presents a transformative shift in the digital landscape. While regulation efforts like the EU AI Act attempt to govern the development and deployment of such technologies, they are unlikely to fully deter malicious actors who disregard compliance. 

Crucially, the challenge extends beyond regulation. Agentic AI will operate within the same digital environments as humans, interacting with the same platforms, interfaces, and services. As machines begin to mimic human behavior online, traditional safeguards like CAPTCHAs become obsolete, demanding that every transaction be authenticated to maintain trust.

In essence, agentic AI doesn’t introduce a new trust problem, but amplifies the existing one, making it both broader and more urgent.

Trust Online Today In Three Questions

At Veriff, we’ve long framed the challenge of online trust around three core questions, which Hubert elaborated on:

Are you who you say you are? This is the bedrock of identity verification (IDV). In the agentic age, IDV remains crucial, particularly for accountability. As Hubert pointed out, the “Cambrian explosion” of digital identity concepts, while promising, also presents significant challenges for businesses trying to support a diverse range of IDs, from physical documents to various digital credentials. This complexity, ironically, drives the need for sophisticated, integrated IDV solutions that go beyond DIY approaches.

Can you be trusted? This delves into the realm of knowing enough about a user to assess their risk profile. For human principals, this involves building a comprehensive customer trust profile, while for agents, it requires establishing a behavioural baseline signature to distinguish normal operation from rogue or fraudulent activity.

Are you a person? And further, are you the person related to the account? This addresses the critical role of liveness and biometric authentication. With the increasing sophistication of deepfakes, liveness checks are becoming even more integral to high-risk transactions like account creation, high-value payments, or account recovery. As Hubert eloquently put it, “Liveness check is therefore the anti-agent agent. It’s going to get busy rapidly.” He emphasized the importance of partnering with providers who have their own robust IP and a vertically integrated approach, where IDV and liveness checks work seamlessly together, creating an “airtight” verification plane.

The Fourth Pillar of Trust in the Agentic Age

However, the advent of agentic AI introduces a crucial fourth question:

Are you authorised? This is perhaps the most significant new challenge brought by agents. As agents interact with various ecosystems, the question of what they are authorized to do – by law, by their developer, by their deployer, and most importantly, by the principal (the human owner) – becomes paramount. While some protocols exist, they are not yet fully equipped for the nuances of agentic AI, requiring dynamic, context-aware authorization and real-time consent mechanisms.

Building a Resilient Infrastructure of Trust

Hubert’s vision is not about reinventing the wheel but rather evolving and augmenting today’s best-in-class trust-building methods. He stressed the importance of a holistic “infrastructure of trust” – not just a collection of techniques, but a cohesive mesh that maintains trust throughout the entire customer lifecycle. This involves:

• Robust Verification: Adapting proven identity and authentication practices to the agentic paradigm.
• Vertical Integration: Combining capabilities like IDV and liveness into a unified, cross-validating system.
• Humans-in-the-Loop: Recognizing the irreplaceable judgment of human oversight to withstand malicious actors and support legitimate agentic AI use.

Finally, Hubert touched upon the critical matter of accountability, introducing Veriff’s new initiative offering refunds for verification mistakes, emphasizing that service providers must demonstrate trust in their own technology. This commitment to accountability will be a key differentiator in a world increasingly reliant on automated trust decisions.

As agentic AI becomes a reality, the identity industry faces its greatest test. But as Veriff’s CPTO made clear in his keynote, what’s required isn’t a reinvention, it’s an evolution. Navigating the complex challenges that the future presents, with thoughtful evolution and robust solutions, we can build a more resilient and scalable infrastructure of trust.