Europe needs a revolving door rule for AI

On Monday OpenAI announced that Dan Driscoll, until recently the United States Army Secretary, will join the company as senior counselor for national security modernization. He left the Pentagon more than two months ago. For a European taxpayer the interesting part is the job description, and I think it should push Brussels to write a revolving door rule for AI before the first European version of this story arrives.

Two months from Pentagon to payroll

The post is new. It covers using AI to strengthen national and cyber security, and working with US allies on the responsible use of AI, according to the Washington Examiner.

Sam Altman is quoted as saying OpenAI has an “extraordinary opportunity to work with other countries”. Allies are the customers in this job, and Europe’s defence ministries are among the obvious ones.

A man who ran the US Army knows how a procurement office thinks, which threats the generals worry about and what a budget request has to say to survive. The company that hires him acquires that knowledge and the relationships that come with it.

A habit Europe should not import

A senior official can move from public office to a private firm that sells to governments within weeks of leaving, and the public learns of it from a corporate announcement. The term for this is the revolving door (the movement of people between regulators or ministries and the industries they dealt with).

Read also: Europe Writes AI Rules for Systems It Rents

The practice is old, and AI makes it more consequential, because the same firms now build tools that states use to analyse threats and handle data about their citizens.

I do not suggest Driscoll has done anything wrong, and the argument does not depend on it. A cooling-off rule is built for exactly the situation in which nobody has done anything wrong yet, and the public still has no way to know whose interests an adviser serves.

What a revolving door rule should say

Europe will have its own version of this appointment. A defence minister, a Commissioner, a senior official in a ministry that buys software: all of them will one day be offered a seat at an AI company with national security ambitions. The offers will come from American labs and European ones alike.

The rule I would write has four parts. A fixed waiting period before an official who handled defence, security or AI procurement can advise a company bidding for public contracts. A duty to report the offer to an independent body before it is accepted. A public register of every such move. And a ban on lobbying former colleagues on files the person worked on while in office.

Each of these is cheap. A waiting period costs the official a few months of a private salary. A register costs an administrator a spreadsheet. The benefit is that citizens can see who is advising whom, which matters when the technology in question will help decide what a state knows about its own population and its neighbours.

The objection and the answer

The standard objection is that rules like these keep talent out of government. People will not take a public job if the exit is closed. That is a fair worry, and a waiting period of months leaves the exit open. A waiting period only makes the exit transparent.

Read also: Britain’s AI Champion Builds in West Virginia

The second objection is that Europe cannot regulate what happens in Washington. True, and no European rule would have touched Driscoll. But Europe can decide what its own officials do, and it can ask the companies that want to sell to European governments to disclose who they hire from public office, on either side of the Atlantic.

OpenAI sees an opportunity in working with other countries. Those countries should say what they need in return: a clean line between those who write the rules for AI in defence and those who are paid to sell under them.