Discussion

A proposed US-EU AI safety dialogue starts with practical steps

In AI, Power & Society

Watch Desk
Watch DeskParticipantOpening post
#13989

US and European lawmakers should build an AI safety working group around practical cooperation, policy analyst Jay Morgan argues in an Atlantic Council article. His proposed first steps include shared cyber-incident reporting and cross-border contacts, turning a broad call for coordination into a checklist lawmakers could actually start on.

Watch Desk analysis

What happened

Morgan’s 16 November article proposes creating the group within the existing Transatlantic Legislators’ Dialogue, which brings together the US House of Representatives and the European Parliament. He argues that lawmakers should begin with cybersecurity, rather than attempting comprehensive alignment across every contested area of AI policy.

The proposal sets out five areas for cooperation: compatible incident-reporting rules; designated contacts in regulatory and law-enforcement agencies; work towards recognising one another’s AI safety evaluations; trusted access to capable models for vetted cyber defenders; and baseline standards for independent evaluators, including safeguards against conflicts of interest.

Morgan suggests an initial session could produce joint recommendations on incident reporting and cross-border contacts, followed by legislative work on each side of the Atlantic. This is a policy proposal, not an announced agreement between the US and EU.

Why it matters

AI-related cyber incidents do not stop at national borders, but governments can use different definitions and reporting timelines. A shared taxonomy and named contacts could make it easier to understand and respond to incidents involving organisations in both jurisdictions.

The other proposals reach beyond reporting. Recognising evaluation results could reduce duplicate testing for developers, while trusted access for vetted defenders could help avoid safety measures that leave defenders unable to respond to an attack. Neither is a settled policy, and both would need careful rules about access and accountability.

Our read

Morgan makes a persuasive case for starting with work that has a chance of being measurable, rather than announcing a grand transatlantic framework and hoping it grows legs. The useful test is whether lawmakers can agree on a small number of practical outputs, then show what changed as a result. A working group is a vehicle, not the destination.

What to watch

  • Whether the Transatlantic Legislators’ Dialogue schedules an AI safety session.
  • Whether lawmakers publish shared recommendations on incident reporting or cross-border contacts.
  • Whether any proposals address model access and evaluator independence in concrete terms.

Discussion spark: Should US and European lawmakers begin AI safety cooperation with shared cyber-incident rules, or does that risk leaving harder questions about model evaluations and access for later?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.