Viable systems, judgement, and AI safety

Rethinking AI safety for agentic systems

Web Summit Lisbon 2026, Stage 13 — AI Summit

Speaker photo

Neil D. Lawrence

Trent.AI and University of Cambridge

AI safety - an accountability problem

Accounting is the numbers. Accountability is the human authority and the judgment.

Agentic debt

devolving authority without judgment undermines accountability and creates agentic debt

Augmented engineers

delegation requires understanding, augmentation supplies understanding

Thanks!

References