States have carved out broad exemptions for national security purposes in existing regional AI legislation – including the EU AI Act and the Council of Europe (CoE) Framework Convention on AI. ECNL did an extensive review of jurisprudence to better understand how national security is defined legally and mapped the instances where it is used in the context of AI. Our research shows that the definition of national security is easily stretched and, as a result, states are invoking it to justify AI deployment in ways that evade regulatory scrutiny, particularly in surveillance and counter-terrorism, but also in other areas.
We used these research findings in our policy advocacy work, helping to address the gap on AI-specific regulation. We took part in the work of the CoE Steering Committee on Human Rights (CDDH) on a new Handbook on AI and Human Rights, a practical guide for European policymakers and public authorities. This is currently the first and only policy framework that clarifies how the existing human rights protections should apply in sensitive areas, such developing and using AI for national security, where oversight is often weakest. Acting as civil society representatives through the Conference of INGOs, we contributed to shaping the chapter on law enforcement and public security, ensuring that real-world AI uses - such as intelligence and identification services, crime prevention and investigation, counter-terrorism and other measures entailing surveillance of people and communications - are assessed against existing human rights standards.
Looking ahead, ECNL will work with civil society organisations, national human rights institutions, local authorities and tech actors to help them embed the Handbook’s guidance in the lifecycle and implementation of AI for national security.
The difference this work made:
- The adopted CDDH Handbook on AI and Human Rights was shaped by civil society expertise and included the recommendations put forward by ECNL.
- The first policy framework is created that addresses the “national security loophole” in AI governance. This helps ensuring that European governments remain accountable to human rights obligations when deploying AI for national security purposes.
- European governments, national human rights institutions and civil society organisations are better equipped in embedding human rights safeguards across the lifecycle of AI systems used for security purposes.
Image: Pauline Wee & DAIR / https://betterimagesofai.org