Global alignment on A.I. is an illusion

COMMENTARY BY Camille Stewart Gloster

For years, policymakers have pursued “global alignment” on artificial intelligence. The phrase appears in summit communiqués, industry pledges and statements from the United Nations. It’s the diplomatic equivalent of a mirage—promising from a distance but evaporating when you get close.

The reality is that AI governance was fractured before it even began, hardwiring fragmentation into the system at speed and scale. The fault lines aren’t just between democracies and authoritarian regimes; they run through alliances once thought unshakable.

The United States has pulled away from Europe’s rights-based, precautionary approach, favoring an industry-led, market-first model. China is doubling down on its techno-authoritarian framework, embedding AI into its surveillance infrastructure and exporting that model abroad, promising international cooperation in AI development and governance. Across the Global Majority, governments are rejecting both paradigms, asserting digital sovereignty, and demanding co-ownership of AI’s benefits and infrastructure.

In this context, “alignment” begins to feel less like a governance strategy and more like diplomatic theater—ritualized language that obscures the widening chasm between regulatory ideologies, enforcement mechanisms, and economic incentives.

But walking away from alignment entirely would be a mistake. Without some common ground, we risk a regulatory free-for-all that accelerates systemic risks, undermines trust, and leaves the most vulnerable with the least protection.

The problem isn’t the goal; it’s the approach. Instead of forcing premature consensus, we should be building interoperable governance zones: modular, purpose-built ecosystems of AI regulation that reflect regional realities while still cooperating, exchanging and converging where it matters most.

Interoperability isn’t a consolation prize. It’s the most realistic scaffolding for future alignment, an architecture that accepts political pluralism but refuses to accept regulatory paralysis. Done right, it would preserve regional agency by allowing countries or blocs to govern based on their own values, economic priorities and risk tolerances, without being coerced into a single governance model that doesn’t fit their realities. It would enable cooperation on shared threats, such as AI-driven cyberattacks, misinformation campaigns, deepfake-enabled election interference and synthetic biology risks, where no one can afford to act alone. And it would facilitate technical interoperability, shared taxonomies, standardized audit protocols, and compatible transparency requirements, and policy interoperability, reciprocal recognition agreements, joint enforcement task forces, and shared crisis-response mechanisms, so that AI safety thresholds in one jurisdiction can be recognized and acted upon in another.

Traditional governance models, especially those under the UN umbrella, rely on consensus. That culture of agreement, while admirable, encourages lowest-common-denominator outcomes: principles broad enough to secure buy-in from nearly all member states but often too vague to be directly enforced. The result is what we’ve seen with other global frameworks: lofty declarations, uneven adoption, and timelines that lag far behind the pace of technological change.

Advancing this model requires government cooperation and industry and civil society coordination happening in tandem. On the government side, regulators can use forums like the Organization for Economic Cooperation and Development, UN-adjacent working groups, and regional alliances to lock in policy interoperability: aligning on shared threat taxonomies, formalizing mutual recognition of AI safety certifications and creating cross-border investigation and crisis-response capabilities. Multilateral technical bodies such as the International Organization for Standardization and the International Telecommunication Union can help set the groundwork for technical interoperability, but governments must translate these voluntary standards into binding obligations backed by enforcement authority.

Meanwhile, industry and civil society must move faster, pushing the frontiers of both technical and policy interoperability in real time. AI developers, infrastructure providers, and governance-focused NGOs can agree on standardized risk assessment methods, audit protocols, and transparency requirements that work across jurisdictions. They can also form enforcement coalitions—revoking access for unsafe models, sharing misuse intelligence across companies, and establishing multi-organization rapid response teams for incidents that cross borders. Civil society actors can bring oversight and public accountability, ensuring these efforts remain transparent and aligned with the public interest.

Of course, interoperability doesn’t erase hard questions. What happens when one zone, say the United States, opts for a hands-off, industry-first model, while another, such as Europe, doubles down on heavy regulation? In the short term, yes, innovation may cluster where rules are looser and capital flows faster. But interoperability can help prevent that divergence from becoming too destabilizing. By requiring even the most permissive zones to recognize minimum safety thresholds and cooperate on cross-border risks, interoperability reduces the race-to-the-bottom dynamic. And history suggests that “move fast and break things” doesn’t scale well in critical infrastructure. Europe’s stricter frameworks may slow some commercial deployment, but they also build trust, resilience, and exportable norms. Over time, innovation follows stability as much as speed, and interoperable governance zones create the conditions for both.

Bodies like ISO, ITU, and the OECD already help lay the groundwork for AI cooperation. But without the connective tissue that turns their convening power into operational interoperability—standards specific enough to be auditable, agreements that carry legal weight, and mechanisms that act in hours, not years—they remain architects without construction crews. Coordinated action between governments, industry, and civil society allows those crews to start building now, each reinforcing the other’s progress.

We already have models for this. Financial regulators coordinate across different economic systems through shared reporting protocols and mutual recognition of risk ratings. Aviation safety agencies exchange accident data, cross-train inspectors, and recognize each other’s certifications. These systems tolerate significant national variation while still achieving a baseline of collective security.

AI needs the same pragmatism but with faster iteration cycles and deeper technical engagement. Imagine a “Basel Accords meets GitHub” approach: the Basel-style element would set minimum capital-like requirements for AI safety, including risk assessments, incident reporting, and operational resilience, while the GitHub-style element would make those frameworks versionable and transparent, enabling governance zones to fork, merge and iterate their approaches without losing compatibility where it matters.

Interoperable governance zones start with fragmentation as the baseline and build connective tissue where incentives align—whether on fighting cybercrime, securing supply chains for AI chips or preventing catastrophic misuse of frontier models. If we wait for a single universal framework, we’ll be waiting as the technology, and its risks, outrun our capacity to shape them. In an era of deepening geopolitical mistrust, cooperation between governments, industry and civil society on both technical and policy interoperability may be the only credible foundation for eventual convergence and the best hope we have of turning the illusion of alignment into something real.

Camille Stewart Gloster, Esq. is the CEO of CAS Strategies, LLC and the former Deputy National Cyber Director for Technology & Ecosystem Security for The White House. In her role, Camille led technology, supply chain, data security, and cyber workforce and education efforts for the Office of the National Cyber Director (ONCD). Camille is a cyber, technology, and national security strategist and policy leader whose career has spanned the private, public and nonprofit sectors. She joined ONCD from Google, where she most recently served as Global Head of Product Security Strategy, and before that as Head of Security Policy and Election Integrity for Google Play and Android.