Auditors Without Power

The audit, as currently practiced in algorithmic accountability, is doing a great deal of cultural work. It is rarely, in any meaningful sense, doing governance work.

This is not a complaint about the people who perform audits. The auditors I have spent time with are conscientious, technically skilled, and clear about what their reports can and cannot say. The problem is structural. We have built an audit aesthetic without building the institutional machinery that historically gave audits force in finance, in medicine, in aviation: tenure, indemnity, subpoena, and a regulator on the other end willing to act.

What audit reports actually do

In the absence of that machinery, an audit report does three things. It records. It performs. It defers. It very rarely binds.

Compliance is a performance, and the report is its script.

This is not new. The accounting scandals of the early 2000s ran on the same substrate: an industry full of careful professionals, producing competent reports, with no consequences attached to the reports themselves. The remedy, eventually, was Sarbanes‑Oxley — not a better template for audits, but a redistribution of legal exposure.

The shape of an alternative

I have written elsewhere about procurement as a more promising site for binding accountability. The short version: contracts have teeth that reports do not. A clause is enforceable; a finding is not.

More to come.