Regulators across Europe are considering their response to AI cybersecurity as Mythos opens up new lines of debate
Key takeaways from this note include:
- Mythos has shifted the policy debate in Europe from a largely theoretical discussion about dual-use cyber capabilities of AI into an urgent governance question about model control and testing.
- In Brussels, the political response will first centre on AI sovereignty and control over frontier systems, with concerns that European regulators and industries remain dependent on arrangements determined by the US and thus are systemically exposed to vulnerabilities.
- The EU AI Office will push for direct testing access to frontier models once its AI Act powers become operational in August – akin to the voluntary UK model; although in practice, the Commission lacks the necessary technical capabilities.
- The Commission and Parliament will also want to explore whether the Cybersecurity Act review could evolve into an additional governance vehicle for AI cyber risks.
- The financial services sector is emerging as the most consequential regulatory battleground, with central banks, prudential regulators, and major firms increasingly seeking independent model testing as concerns grow that AI-enabled cyber capabilities could become a systemic financial stability risk.
Fill in this form to read the full note >>>