Ethics Are the Scaffolding, Law Is the Building

FATIMA Roumate is a Full Professor of International Economic Law and Artificial Intelligence at Mohammed V University in Rabat, Morocco, and holds an honorary doctorate from UNIFE in Peru. In 2020, she was nominated by UNESCO as a member of its ad hoc Expert Group on AI ethics.

The Confluence: How did you first get drawn into the governance and ethics of AI as a research and policy focus?

Dr. Roumate: I began working on my specialisation in 1997, starting with graduate research on the role of technology investment in Moroccan development and later a PhD thesis on the challenges of Moroccan technology policy. In 2013, I began focusing specifically on the link between AI and international relations, publishing on AI’s challenges and risks and a country’s ability to control its own technology and data.

In 2020, I was nominated by UNESCO as a member of its ad hoc Expert Group on AI ethics, and the same year joined its Working Group on Information Ethics. I am also a member of the ITU’s AI for Good Impact Initiative Steering Committee and of its Focus Group on AI and Environmental Efficiency.

In 2010 I founded the International Institute of Scientific Research, and I later founded the Global Network of AI and International Society. I continue as an expert member of several working groups within international organisations, and with think tanks in the USA and Europe, and I write in Arabic, French and English: my edited book Artificial Intelligence and Digital Diplomacy: Challenges and Opportunities and my 2023 Springer Nature book Artificial Intelligence, Higher Education and Scientific Research: Future Development both grew out of this work.

The Confluence: For a reader who assumes international law can simply be updated to handle AI, where does that assumption break down in practice?

Dr. Roumate: That assumption stems from a fundamental misunderstanding of what AI represents: not a novel tool to be appended to existing treaties, but a structural transformation that challenges the foundations of the Westphalian order [the system, dating back to the Peace of Westphalia in 1648, in which sovereign states are the only full players in international affairs] itself. It breaks down across several dimensions.

First, international law presumes the sovereign state is the exclusive subject of international public law, but AI has allowed transnational Big Tech corporations to emerge as powerful non-state actors that own the foundational infrastructure of the modern world, a kind of “cable sovereignty” that rivals states directly [companies such as Google and Meta now own or part-own many of the undersea cables that carry the world’s internet traffic].

Second, the mechanics of diplomacy, multilateral negotiations, and ratifications are notoriously slow, while AI develops exponentially, a temporal disconnect aggravated by states’ own confusion over concepts like “data sovereignty” [a country’s claim to control data generated within its borders by its citizens]. We saw this friction during COVID-19, when tensions over data sovereignty between the US and China complicated the free flow of scientific data.

Third, patching existing rules is not enough: we need entirely new doctrines, such as “international psychological security” [protecting populations from foreign campaigns designed to manipulate what people believe, such as AI-generated fake news], to address AI-driven disinformation, backed by a permanent UN scientific body of legal scholars and engineers working in unison.

Fourth, international law has always assumed a human interpreter, and the proliferation of autonomous systems now forces a reassessment of legal personhood [who, or what, can hold rights and be held responsible in law] itself.

Finally, the assumption collapses most completely around Lethal Autonomous Weapons Systems, “human-out-of-the-loop” systems that select and engage targets without meaningful human intervention, undermining the UN Charter’s core assumptions about the use of force and making it unclear whether a machine can reliably distinguish combatants from civilians.

Confronting this reality requires a profound structural overhaul of global governance and an entirely new class of international instruments for a world where humans are no longer the sole actors.

The Confluence: Is there a specific AI governance debate right now where you think the public conversation is missing something important?

Dr. Roumate: The global debate is characterised by a significant conceptual vacuum: the international community has failed to agree on a common definition of “AI governance” itself, and the term has been co-opted by competing actors.

Big Tech corporations treat governance as a synonym for “controlling AI,” meaning managing operational risk and reliability to maximise profit, framing it as an internal company discipline that can substitute for the traditional power of law. Codes established by corporations are not law but are increasingly treated as such, even though algorithmic codes cannot be superior to national constitutions or international law.

Meanwhile, states recognising that technological sovereignty [a country’s ability to build and control the technology it depends on] is the only barrier against foreign interference have mostly retreated into isolated, national-level policy, which is fundamentally flawed since AI is inherently borderless: an isolated statute passed by a single parliament cannot govern a decentralised, transnational AI ecosystem, and tech monopolies simply shift operations to jurisdictions with the weakest oversight.

AI governance cannot be outsourced to Big Tech; it must be anchored in public and international law, since protecting human rights remains the mission of the state, while corporations are driven by market and profit.

The Confluence: What would you say to someone who’s sceptical that ethics frameworks, as opposed to hard regulation, can constrain how AI gets built and used?

Dr. Roumate: Their scepticism is justified, but it misses a nuance: ethics and hard law are not mutually exclusive, they are sequential. Ethical frameworks serve as an architectural blueprint for future binding instruments, offering a vocabulary and baseline consensus before states can code expectations into enforceable norms.

But AI ethics are essential and insufficient. An ethics checklist cannot compel an algorithm to respect the Geneva Conventions [the treaties setting the rules of war, including the protection of civilians] when a state is locked in a military conflict; when the survival of a nation or the profit of a corporation is on the line, ethical guidance is the first thing abandoned.

Ethical frameworks are effective at guiding responsible design in the laboratory, but they lack the teeth to constrain geopolitical ambition and corporate greed on the global stage. Ethics are the scaffolding; hard law is the building. To truly protect human rights, frameworks such as the UNESCO Recommendation on the Ethics of Artificial Intelligence [adopted by UNESCO’s member states in 2021, but not legally binding] must be codified into binding international and regional instruments, backed by state sovereignty.

The Confluence: If you could get policymakers to understand one thing about AI governance that’s currently underappreciated, what would it be?

Dr. Roumate: AI governance should not be reduced to controlling AI; it should refer to a legal, institutional, and financial framework that guides AI innovation for the good and well-being of all, and because AI’s challenges transcend national borders, it must be led by international organisations and their member states, not treated as a matter for isolated national statutes.

That said, these frameworks cannot be dictated from an ivory tower: they must be grounded in a multisectoral, multidisciplinary approach that includes Big Tech, academia, and civil society in their implementation, even as states retain ultimate legal authority. This is not an abdication of state power but a modernisation of diplomacy for a world where AI may no longer be a tool but an agent, and potentially a competitor.

Above all, policymakers should understand that AI, and AI surveillance especially, is a trap: used to control citizens without protecting their rights, it risks costing governments the trust that is the foundation of their own legitimacy. AI governance must be grounded in the balance between innovation and the protection of human rights, or the state will lose its legitimacy and independence to the great powers and non-state actors who already hold technological sovereignty.