Atlantic Council urges US and EU lawmakers to coordinate on AI safety

US lawmakers and the European Parliament should establish a formal dialogue on artificial intelligence safety, beginning with cybersecurity, the Atlantic Council argues. The proposed cooperation would seek compatible rules between the United States and the European Union while avoiding an attempt to align every aspect of AI regulation at once.
- The Atlantic Council calls for a US-EU legislative dialogue on AI safety.
- It proposes a joint working group within the Transatlantic Legislators' Dialogue.
- Cybersecurity, incident reporting and cross-border contacts would come first.
- The report outlines five potential areas for practical cooperation.
- It recommends narrow, measurable steps rather than immediate regulatory alignment.
A proposed legislative channel
The Atlantic Council says international AI safety discussions have focused heavily on the United States and China, but argues that Europe must be included in any workable global governance system. The European Union's common market and its regulatory role make transatlantic coordination important, the analysis says.
Rather than relying only on executive action, the report recommends that Washington and Brussels create a joint AI safety working group within the existing Transatlantic Legislators' Dialogue, or TLD. The forum brings together the US House of Representatives and the European Parliament, meeting twice a year since 1999 in alternating locations in the United States and Europe.
The group could begin with cybersecurity measures, where the report identifies an acute risk from AI-enabled attacks, and expand its work over time. The TLD's flexible structure could support an AI safety-focused session when the new Congress convenes next year, according to the Atlantic Council.
Five areas for early cooperation
The report proposes five short- to medium-term deliverables. The first is interoperable incident reporting, including a common taxonomy and compatible deadlines for reporting cybersecurity incidents. It points to the EU's AI Act and NIS2 Directive, as well as the US Cyber Incident Reporting for Critical Infrastructure Act, while acknowledging that the existing frameworks differ in their triggers and coverage.
- Common incident reporting definitions, timelines and procedures
- Designated regulatory and law enforcement contacts on both sides of the Atlantic
- Coordination between US and EU government testing and evaluation programs
- Trusted access to more capable frontier models for vetted cybersecurity defenders
- Baseline accreditation and independence standards for third-party evaluators
The proposed evaluator cooperation would involve the US Center for Advancing Innovation and Standards for Super Intelligence and the EU AI Office. The report also says the United Kingdom's AI Security Institute should ideally participate, with the goal of reducing duplicate safety evaluations for developers operating in major markets.
Why a narrow start matters
The Atlantic Council cautions that previous transatlantic discussions have struggled when either side believed the other was trying to export its regulatory model. Dialogues have also been criticized as producing too few concrete results for constituents.
For that reason, the report recommends starting with narrowly defined outcomes. An initial AI-focused TLD meeting could produce joint recommendations on incident reporting and cross-border points of contact. Those recommendations could then inform draft legislation in the House and a legislative initiative report in the European Parliament, with Parliament potentially asking the European Commission to propose legislation aligned with the shared framework.
The approach would operate across different US and EU policy systems. US AI safety policy is described as a mix of voluntary commitments, executive actions and state legislation, while the EU uses the AI Act as its legal framework. The report says amendments adopted through the Digital Omnibus, which entered into force in July 2026, changed implementation timelines and compliance requirements, leaving issues unresolved.
For users and AI companies, the proposed dialogue could eventually mean more compatible incident reporting, testing and emergency-response expectations across the two markets. The immediate proposal is more limited: build cooperation around cybersecurity first, demonstrate practical results, and use those results to decide whether broader AI safety coordination is workable.
Atlantic Council (AI): Why Congress and the European Parliament need an AI safety dialogue


