How governments worldwide are shaping what AI can say about politics
Some governments use legitimate AI concerns as a pretext to control political discussion, particularly chatbot responses about those in power.
As the U.S. midterms approach, about one in five voters have turned to AI chatbots for election and political information, according to a recent survey. However, many voters may be unaware that governments worldwide are increasingly dictating what these chatbots can say. This is because tech companies often implement the same policies globally, meaning American voters' responses could be influenced by both domestic and international governments.
Researchers at a nonpartisan think tank at Vanderbilt University examined how 12 countries and regions regulate artificial intelligence. They found that governments frequently rely on existing laws, such as national security or anti-state propaganda statutes, rather than creating new AI-specific regulations to control what AI models say.
While addressing these harms is crucial, some governments misuse these concerns as a pretext to control political discussions, especially concerning chatbot responses about individuals in power.
For instance, in 2024, an Indian journalist asked Google's Gemini chatbot about India's prime minister, Narendra Modi's policies, and received a response suggesting some experts view his leadership as fascist. This prompted the Indian government to require official permission for "unreliable" AI models to be released. Although the requirement was later dropped, it illustrates how terms like "unreliable" can be weaponized for political gain.
China mandates that AI models adhere to "socialist core values," while Turkey banned Grok, the chatbot on social platform X, for allegedly insulting President Erdoğan and religious values. In Vietnam, two activists living in Germany were recently sentenced to 17 years in prison for spreading anti-state material by using AI to impersonate Vietnamese leaders through fake voices and images.
Many of these government measures target people within their borders, but some extend beyond national boundaries. The European Union's AI Act, implemented in 2024, mandates that companies offering AI tools in Europe, irrespective of their location, mark AI-generated content. This global impact is part of the Brussels Effect, where EU rules influence other jurisdictions.
The EU also rated ChatGPT under the strictest tier of its Digital Services Act, requiring OpenAI to assess and reduce systemic risks, including election manipulation through disinformation. A study by the Dutch Data Protection Authority found that political discourse on AI chatbots predominantly favored one side of the middle ground during the 2025 Dutch election, urging providers to prevent their systems from being used for political campaigning.
Domestically, the United States has its own concerns. President Trump's 2025 executive order on "Preventing Woke AI in the Federal Government" demands that AI models used by the government be "truth-seeking," ideologically neutral, and not biased toward concepts like diversity, equity, and inclusion. However, the definition of "truth-seeking" can vary widely.
For example, the president recently banned three news outlets for reporting "FICTION and LIES," and the Federal Trade Commission has warned that AI companies steering answers with undisclosed ideological biases might violate consumer protection laws. A 2025 Pew Research Center survey revealed that eight in ten Americans believe Republican and Democratic voters disagree on basic facts.
Thus, government mandates to seek the truth in AI responses are often influenced by ideological beliefs, particularly on political issues.
Written by urgent.news from The Conversation's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
