Unverifiable quality - the Al lemons problem - is about to become a deep-technology problem rather than a product
problem. Part 1 of this article argued that Al's biggest commercial problem is not intelligence but credibility.
When Al is surfacing information (for example via an LLM interface) a user worries that the machine may be wrong; but
in that case its outputs can still be checked. The problem is merely whether we can trust an Al-generated answer. This is
how Al works in old kinds of firms: firms that employ mainly people use Al to improve operational efficiency and use people to check the answers. New kinds of firms, founded on machine intelligence and labour, require us to consider whether we can trust an Al-generated action; in the agentic economy a machine may be wrong or unauthorised and still act.
Both parts of this article consider trust but, in this part, concerning the machine (or agentic) economy, we look at trust
through automated verification: systems that test machine action, machine authority, and machine evidence, and thereby develop legally informed systems that test the boundaries within which agents act.Old kinds of firms employ mainly people; new kinds of firms employ mainly machines.
26 JUL 2026
CMS reviews key market developments in the banking sector
26 JUL 2026
CMS reviews key market developments in the banking sector
25 JUN 2026
CMS reviews key market developments in the banking sector
22 MAY 2026
CMS reviews key market developments in the banking sector
4 MAY 2026
CMS reviews key market developments in the banking sector
12 APR 2026
CMS reviews key market developments in the banking sector
27 FEB 2026
Jason Blick considers the implications of the SAFE regulation for investors and creditors including “third country” entities.
27 FEB 2026
CMS reviews key market developments in the banking sector
8 FEB 2026
CMS reviews key market developments in the banking sector
10 JAN 2026