CitizenGO

The first Global Dialogue would be a success if it establishes one non-negotiable principle: AI must never become an instrument of censorship by governments or international institutions. The greatest threat posed by AI governance today is not the absence of regulation but the risk of over-regulation, where governments use "safety" and "trustworthiness" as pretexts to control what citizens can say, read, and think. A successful outcome would be a binding commitment that no AI governance framework shall empower any government or international body to determine what constitutes acceptable speech. The Dialogue must reject the growing trend of governments pressuring platforms to use AI to suppress lawful expression labeled as "disinformation" or "hate speech," terms that are increasingly weaponized against political dissent, religious expression, and moral convictions that challenge the ruling consensus. Success means more freedom, not less.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Safe, secure and trustworthy AI

Our selection is driven by one overriding concern: the use of AI as a tool for public censorship. The most fundamental human right in the digital age is freedom of expression, and AI-powered content moderation, deployed under government pressure or mandate, is the single greatest threat to that right today. CitizenGO represents over 20 million citizens in 50 countries who have directly experienced how AI systems silence lawful speech on contested issues like the protection of life, family, and religious freedom. Transparency is essential because censorship thrives in opacity: citizens cannot challenge what they cannot see. When governments compel platforms to deploy AI filters against "harmful content," the public must know exactly what criteria are applied and who defined them. The social and cultural implications of AI matter because these systems are being used to enforce a single worldview globally, punishing dissent from approved narratives. "Safe AI" must mean safe from government abuse, not safe from ideas that make authorities uncomfortable.

The most dangerous gap in current AI governance discussions is the normalization of state-directed censorship through AI. Governments worldwide, including in democratic nations, are building legal and institutional frameworks that require or incentivize platforms to use AI to suppress speech they deem "dangerous," "misleading," or "hateful." The EU Digital Services Act, Brazil's regulatory proposals, and similar initiatives risk creating a global infrastructure for automated censorship with no meaningful accountability. This is not a hypothetical threat: it is happening now. A second critical gap is the role of international organizations themselves in promoting AI-driven content moderation. When UN agencies define "disinformation" and then pressure platforms to suppress it using AI, they are acting as censors with no democratic mandate. The AI Dialogue must confront this honestly rather than reinforcing it. Any governance framework that gives governments or international bodies the power to define what AI should suppress is a framework for tyranny, regardless of the intentions behind it.

The primary impact on civil society is not from the absence of governance but from its excess. In Europe, the Digital Services Act is creating a regime where platforms preemptively censor lawful speech using AI to avoid regulatory penalties, a system of privatized censorship driven by government coercion. CitizenGO campaigns have been suppressed across Spain, France, Germany, and Latin America by algorithms responding to government and institutional pressure, not to any violation of law. In the Global South, governments are adopting European-style "content moderation" frameworks that hand authoritarian regimes a sophisticated tool for silencing opposition under the guise of fighting "disinformation." The opportunity is enormous: if the AI Dialogue establishes that AI governance must protect citizens from censorship rather than enable it, it could prevent the construction of a global automated censorship infrastructure. The principle must be clear: no AI system should be used, whether by governments or at their direction, to suppress lawful expression. The role of governance is to protect freedom, not to manage what people are allowed to say.

The AI Dialogue should serve as a bulwark against the global trend of governments using AI governance as a vehicle for censorship. Its most important contribution would be establishing an international norm that AI governance frameworks must never empower any state or institution to use AI for suppressing lawful expression. The Dialogue should create a watchdog function, monitoring and publicly denouncing cases where governments use AI regulation to silence dissent, religious speech, or political opposition. International cooperation on AI should focus on preventing a race to the bottom where countries compete to build the most effective censorship infrastructure. The Dialogue must resist pressure from governments that seek international legitimacy for their domestic censorship regimes. It should champion the principle that the answer to speech one disagrees with is more speech, not algorithmic suppression, and that no international consensus can override the fundamental right of every human being to speak, believe, and advocate freely.

• The Dialogue should look critically at the US First Amendment tradition, which establishes that government-directed suppression of speech, including through private intermediaries, violates fundamental rights. Recent US court rulings on government coercion of platforms to censor content provide valuable jurisprudence. The UN Special Rapporteur on freedom of opinion and expression has produced important work warning against the dangers of content moderation, and this should be central to the Dialogue. The Dialogue should explicitly critique initiatives that have enabled censorship: the EU Digital Services Act's systemic risk framework, which incentivizes over-removal of content
• the Christchurch Call, which has been used to justify expansive AI-driven suppression beyond its original scope
• and various "countering disinformation" programs by international organizations that define legitimate political debate as mis/disinformation. The added value of this Dialogue should be to serve as a corrective, bringing voices that have been censored into the room and forcing honest discussion about whether current AI governance trends are protecting rights or destroying them.

The most important structural reform is ensuring that the AI Dialogue does not replicate the closed, ideologically homogeneous stakeholder model that dominates current tech governance. Organizations that have been targets of AI-driven censorship must have a seat at the table, not as token participants but as central voices. The Dialogue should include a dedicated track on "AI and censorship" where affected parties present documented cases. Governments that are building censorship infrastructure should be challenged publicly, not given platforms to present their frameworks as best practices. Tech companies should be required to participate in adversarial sessions where civil society can question their content moderation practices under oath or formal commitment. The format must prevent capture by the usual coalition of governments and aligned NGOs that treat AI governance as an opportunity to expand state control over speech.

The most silenced voices in AI governance discussions are precisely those being silenced by AI: religious communities, pro-life organizations, pro-family advocates, political dissidents, and citizens who challenge progressive orthodoxy on issues from gender ideology to immigration. These groups represent billions of people worldwide but are systematically excluded from "multistakeholder" processes that claim to represent civil society. Faith-based organizations, which serve more people globally than most governments, are virtually absent. Conservative and traditionalist civil society from Africa, Latin America, the Middle East, and Asia is ignored in favor of Western-funded NGOs that mirror the values of the tech industry and donor governments. Including these voices requires abandoning ideological gatekeeping in stakeholder selection, actively recruiting organizations that have been censored by AI systems, and providing financial support for participation from the Global South, where travel to Geneva or New York is prohibitively expensive.

The most powerful format would be live "censorship audits" where civil society organizations from different ideological perspectives demonstrate in real-time how major AI systems treat their content differently, making the bias visible and undeniable. A public testimony track where citizens and organizations present documented cases of AI-driven censorship would ground the discussion in reality rather than abstractions. An open online platform collecting censorship reports from around the world would build an evidence base that governments cannot ignore. The Dialogue should also include structured debates between governments that promote AI content regulation and civil society organizations that have been harmed by it, forcing direct confrontation rather than parallel monologues. Regional preparatory consultations must include genuine ideological diversity, not just geographic diversity.

The strongest models for AI governance are those that protect citizens from censorship rather than enabling it. The US First Amendment framework, reinforced by recent federal court decisions limiting government coercion of platform censorship, provides the gold standard. State-level laws in Texas and Florida prohibiting viewpoint-based censorship by platforms point the direction for international norms. Argentina's constitutional tradition of prohibiting prior censorship offers a Latin American model. Poland's proposed legislation giving citizens the right to appeal AI-driven content removal established an important principle, even if implementation is debated. On the negative side, the EU Digital Services Act and Brazil's "fake news" bill represent cautionary examples of how well-intentioned regulation creates infrastructure for state censorship. The most important governance practice is the simplest: governments should not be in the business of deciding what speech AI systems should suppress. Any framework that gives public authorities that power will inevitably be abused, regardless of safeguards. The best AI governance is one that maximizes human freedom and minimizes government control over expression.