AI agents are moving beyond generating content. They can now access systems, retrieve information, communicate with people, trigger processes and increasingly make decisions on our behalf. But giving an AI the technical ability — or even the permission — to act does not necessarily give it the authority to make every decision that follows.
In this episode of Innovating Trust, Steve Atkins puts an AI agent “on the stand” to examine where capability ends and responsibility begins. A seemingly simple instruction to arrange a podcast interview quickly becomes a chain of access, permissions, judgement and delegated authority. If the AI reads correspondence, changes a calendar, communicates externally or decides which meeting matters more, who actually authorised those decisions — and who remains accountable when something goes wrong?
The discussion also looks at AI provenance and watermarking as Europe attempts to establish clearer links between machine-generated content and the people or organisations that stand behind it. The same principle becomes even more important as AI moves from creating information to taking action.
The central argument is simple: identity tells us which system acted, permissions tell us what it could access, and audit trails tell us what it did. Digital trust requires something more — a clear chain of authority and accountability.
As AI becomes more autonomous, responsibility does not disappear. It becomes more important.
The Quantum Space: Innovating Trust is recognised by MillionPodcasts as one of the Top 20 Cryptography Podcasts : view the list here
Listen to ‘Innovating Trust’ on your regular Podcast platform;
Leave a Reply