In Geneva, the World Can Anchor AI Governance in Free Expression

The United Nations is convening its Global Dialogue on Artificial Intelligence Governance in Geneva July 6-7, and the stakes are high: what rights will anchor the future of how the world governs AI? Generative AI systems mediate, filter, and generate the information we encounter every day. They are, at their core, a technology of expression and access to information. How we govern them will have downstream effects on what people can say, seek, and know. Geneva is where we can get that right.

A foundation worth building on

The past several years produced exceptional progress on examining the interactions between generative AI and rights. The Paris AI Action Summit in February 2025 produced a declaration signed by 58 sovereign states that anchored AI governance in human rights and sustainability. Notably, the US and UK declined to join — a fracture in rights-focused alignment by AI powerhouses that contributes to the urgency in Geneva.

Despite that setback, this rights-centered commitment was reinforced months later at the regional organizational level. In October 2025, the UN Special Rapporteur on freedom of expression, alongside counterparts from the Organization for Security and Co-operation in Europe (OSCE), the Organization of American States (OAS), and the African Commission on Human and Peoples’ Rights (ACHPR), issued a joint declaration affirming that freedom of expression must be embedded throughout the AI lifecycle, from design and training to deployment.

These expert-level commitments reflected obligations towards freedom of expression and access to information UN Member States had already adopted by consensus in 2024 through the Global Digital Compact. The Compact committed governments to protecting freedom of expression and access to information when mitigating potential harms and to refrain from information restrictions inconsistent with international law, a meaningful commitment in a universal instrument. It was the Global Digital Compact that established the upcoming Global Dialogue on AI Governance. Even the earliest intergovernmental standard on AI had already embedded this logic, and the 2024 revision to the OECD AI Principles conditioned information integrity measures on respect for freedom of expression.

Together, these instruments establish a baseline that governments have already agreed to. Geneva can build on and strengthen this baseline, ensuring that free expression and access to information are robustly protected in the AI era.

The New Delhi Declaration: an opportunity not taken

This year’s India AI Impact Summit was a missed opportunity for human rights and, in particular, freedom of expression. The international community fell short of building AI governance on a human rights foundation. The New Delhi Declaration, endorsed by more than 90 countries and international organizations, called for “industry-led voluntary measures” and described AI’s purpose as advancing “social good,” without accountability mechanisms or rights standards attached to that phrase.

Before the summit, a keyword analysis of its agenda found “innovation,” “growth,” “efficiency,” and “productivity” dominating the program, while “human rights,” “accountability,” and “surveillance” were largely absent from high-level plenary titles.

The Dialogue's human rights cluster and its independent scientific panel provide the structural anchors that the last summit’s voluntary framework deliberately avoided. This gives Geneva the opportunity to correct what was left behind in New Delhi by reaffirming the integral role of human rights in AI governance.

Why freedom of expression Is central to AI

Generative AI systems now reach over a billion people globally, and information-seeking became the leading use-case of generative AI in 2025 across six countries (Argentina, Denmark, France, Japan, the UK, and the US). Whether or not you directly use a chatbot, you likely receive an AI-generated summary every time you search online through services like Google AI Overviews or Bing Copilot Search. That reach makes AI governance inseparable from the governance of expression itself.

International human rights law provides a solid basis on which to build AI governance in the new era. Article 19 of the International Covenant on Civil and Political Rights, the main instrument protecting freedom of expression globally, protects the right to “seek, receive and impart information and ideas […] regardless of frontiers” through any media. This forward-looking wording provides a solid basis to protect AI-generated expression and information that the international community can rely on.

Article 19 should be top of mind to all those participating in the Global Dialogue this summer. Governments have often invoked information accuracy and public order to restrict sociopolitical expression online, endangering political speech, government criticism, journalism, and advocacy. While information integrity is a genuine concern, there is a danger in invoking it without the necessary counterbalance of Article 19's strict requirements in AI governance. This tripartite test dictates that any restriction must be provided for in law, pursue a legitimate aim, and be necessary to protect a legitimate aim.

Article 19 is also a helpful tool because it enables policymakers to tackle real harms, like non-consensual intimate imagery and child sexual abuse material, while protecting the tenets of democracy.

Why rights protections matter in practice

Over the last three years, it has become clear that the importance of protecting freedom of expression and access to information extends beyond theory. The examples of free speech abuses under vague safety arguments have been too numerous.

In India, when Google’s Gemini generated responses in February 2024 describing the prime minister’s policies in terms some experts characterized as fascist, the government accused Google of violating IT Rules and issued a formal advisory requiring pre-clearance for AI model launches. The message to AI developers was unmistakable: outputs that displease the government create legal exposure. India’s three-hour removal rule for AI-generated content, introduced in early 2026, compounded that pressure. Platforms must now remove flagged unlawful content within three hours of a government notification. The rule draws no distinction between political satire, journalism, and genuinely harmful material, and it reaches political opponents, journalists, ordinary users, and civil society organizations alike. India’s digital censorship infrastructure saw direct content blocking orders more than double to over 24,000 in 2025. The AI removal rule extends that infrastructure into a new medium. However, the use of public order as grounds for restricting AI outputs is not unique to India.

In Turkey, a court blocked access to Grok content in July 2025 after the xAI chatbot generated outputs deemed insulting to President Recep Tayyip Erdoğan. The stated rationale was to protect public order, precisely the kind of rationale that Article 19’s proportionality requirements exist to constrain.

These pressures are not confined to governments with established records of censorship. In the US, domestic debates over neutrality in AI have complicated a principled international stance on freedom of expression. The European Union’s AI Act and Digital Services Act contain systemic risk provisions broad enough to function as speech restrictions, with national regulators already criticizing chatbots for producing “biased” political outputs, which is a pattern worth keeping an eye on.

Most consequentially for Geneva, China's approach is not merely domestic. In July of last year, China released its Global AI Governance Action Plan, invoking “public good” and “safety” as governing principles. Unsurprisingly, China’s domestic AI regime requires adherence to “core socialist values” and restricts outputs that may challenge state authority or social order. This framework is being actively promoted in the same multilateral forums whose outcomes will determine whether freedom of expression is protected in global AI governance.

What the Global Dialogue can achieve

The UNESCO Recommendation on the Ethics of AI of 2021 provides a template on what rights-based AI governance can look like, providing inspiration for what the results of the Global Dialogue may look like. The UNESCO Recommendation requires any limitation on rights to follow the tripartite test (legality, legitimacy, and proportionality), bars AI systems from being used for social scoring or mass surveillance, and requires member states to ensure that AI actors respect rights in the AI lifecycle. This represents a model worth returning to. Building on its foundation, the Global Dialogue can go further: holding freedom of expression and access to information as an operative standard and applying the Article 19 tripartite test for AI content governance.

The Global Dialogue’s human rights cluster and the Independent International Scientific Panel can operationalize commitments that UN Member States have already made. As we argued in our submission to the Global Dialogue’s call for input, we propose:

• The Co-Chairs' Summary and any working group mandates adopted at Geneva should name freedom of expression and access to information explicitly — closely following the Global Digital Compact's commitments — with civil society and independent experts participating substantively in the process.
• The Scientific Panel should assess the latest capabilities of AI-enabled surveillance and content authenticity technology in order to fully assess the potential harms to freedom of expression, alongside privacy.
• The Dialogue should affirm that Article 19’s requirements of legality, legitimacy, and necessity are consistently held throughout the conversation around AI governance, particularly regarding information integrity measures and any speech restrictions.

AI is the information technology of the future, and it deserves governance that gives careful consideration to rights protection. Geneva is where rights-respecting best practices can be made global, breaching national and regional divides.