Who owns the company—and who can see the answer?

Romain Liot moderates Thom Townsend (Open Ownership) and Wendy Thomas (EITI). The topic is data about who ultimately owns and controls companies, not ownership of the dataset itself, though data-access rights become central. Thom defines beneficial ownership in terms of natural persons behind legal entities and explains why corporate chains can obscure control. Historical, legal and geopolitical descriptions are attributed claims, not separately verified findings.

EITI describes annual country reports and granular extractive data, periodic validation and less intensive supporting-company assessment. Wendy explicitly says EITI does not centrally collect granular beneficial-ownership records; it monitors implementation and makes reports and assessments available. Its tracker is not a real-time ownership registry. The transparency theory of change seeks fairer public benefit and reduced corruption, but the audience's question about whose behaviour changes receives only a broad account. A newly expanded monitoring function is described; no causal impact estimates are presented.

Open Ownership previously aggregated public records into a common data standard. Thom describes retreat from unfettered public access, legitimate-interest systems with unclear reuse terms, and an API change that broke pipelines. Keeping republished personally identifiable records current was necessary to its legal position, and the cost and risk became unsustainable. This is a particularly concrete negative case: an available API and nominal access do not establish maintainable, legally usable infrastructure. The reported two-week remapping effort and cloud cost are organisation-specific, not a generic unit-cost estimate.

The access debate is not resolved by asserting that transparency is good. Thom recognises the privacy logic of a court ruling while disputing practical consequences and some risk narratives. His lack of knowledge of kidnapping cases cannot establish that the risk is zero. Public-interest investigation, privacy, institutional capacity and clear licensing must remain separate. Reported examples in Beirut and Ghana illustrate how records can give investigators or advocates a lead; they do not establish completed accountability or broad causal effects. The Zambia aid/extractives example is a geopolitical claim that has not been independently verified here.

An audience member proposes federated self-hosting from a soil-data context. Thom responds that state disclosure systems and personally identifiable data retain legal constraints under federation. This is a direct boundary condition to the event's enthusiasm for distributed architecture. The practical data standard is described as linked person, entity and ownership/control statements, with persistent identifiers needed to disambiguate roles across time and languages. Technical simplicity at the schema layer coexists with institutional complexity in access and maintenance.

The close is cautiously optimistic about policy adoption and forthcoming research, while admitting that evidence alone does not ensure lasting change. The anticipated studies are not completed findings. Commercial secrecy, privacy, political obstruction and ordinary administrative incapacity may all constrain transparency; the speaker's interpretation that opposition proves success is not itself evidence of success.