How governments worldwide are shaping what AI can say about politics

As the U.S. midterms approach, a recent survey indicates that roughly one in five voters has sought information about elections and politics from AI chatbots. However, many voters may be unaware that governments worldwide are actively influencing the answers these chatbots provide. This can occur due to the fact that many tech companies apply the same policies across the globe, leading to potential foreign influences on the information American voters receive.

Researchers at a nonpartisan think tank at Vanderbilt University have conducted studies on how 12 countries and regions regulate artificial intelligence. Their findings reveal that these governments often use existing rules, such as national security or anti-state propaganda laws, rather than creating new AI laws to control the content AI models can generate.

While it is important to address harmful uses of AI, such as cyberattacks or the creation of sexualized images of children, some governments are using genuine concerns like these as a pretext to control political discussions, particularly concerning chatbot responses about those in power.

For example, in India, a journalist asked Google's Gemini chatbot about the policies of Prime Minister Narendra Modi. When the chatbot responded that some experts say Modi's policies are fascist, the Indian government required official permission for releasing "unreliable" AI models. Although this requirement was later dropped, it demonstrated how terms like "unreliable" or "under-testing" could be manipulated for political purposes.

China requires AI models to adhere to "socialist core values," and in Turkey, a court ordered the blocking of Grok, a chatbot in the X social platform, for allegedly insulting President Recep Tayyip Erdoğan and religious values. Turkish law scholar Yaman Akdeniz identifies the Turkish rulings as among the earliest documented cases of state censorship of chatbots.

Vietnam shows how content controls can impact dissidents, even those living abroad. In December 2025, two Vietnamese activists residing in Germany were sentenced to 17 years each in prison for spreading anti-state material using AI to fake voices and images of Vietnamese leaders. This highlights how political criticism can lead to prosecutions under such content controls. Some of these government measures only impact individuals within their countries, while others have a global reach.

The European Union's AI Act, adopted in 2024, mandates that companies offering AI tools in Europe, regardless of their location, must mark AI-generated content. Companies like Anthropic, which lack reliable methods to apply these markings regionally, therefore apply them globally. This approach exemplifies the Brussels Effect, where rules adopted within the EU have a widespread impact.

The EU also influences AI conversation by placing ChatGPT under the strictest tier of its Digital Services Act, requiring OpenAI to assess and mitigate system risks, including those related to elections and disinformation.

These regulations can be utilized to pressure platforms. In 2024, Thierry Breton, then the EU's top digital rules enforcer, cautioned Elon Musk before his interview with then-presidential candidate Donald Trump on X, warning against the spread of "harmful content." The threats to Americans' access to information are not solely foreign in origin.

In 2025, President Trump issued an executive order on "Preventing Woke AI in the Federal Government," requiring the AI models the government purchases to be truthful, ideologically neutral, and not favoring ideas like diversity, equity, and inclusion. While the term "truth-seeking" may appear neutral, it can hold different meanings for various people.

For instance, President Trump recently banned three news outlets from the White House for reporting "fiction and lies" and dismissed media polls as "fake." The Federal Trade Commission has warned that AI companies steering answers toward undisclosed ideological or biased objectives might violate consumer protection laws. A 2025 Pew Research Center survey found that eight in ten Americans believe Republican and Democratic voters cannot agree on basic facts.

This leads to the conclusion that government mandates to seek truth in AI often stem from ideological beliefs, especially on political issues.

Additionally, several states have enacted laws restricting AI-generated political deepfakes. California and Hawaii, for example, have prohibited the sharing of deceptive AI content about candidates in the months leading up to an election. However, these laws have been blocked by U.S. courts on First Amendment grounds.

For voters, it is crucial to treat AI chatbot responses about elections as a starting point. Remember that governments, both foreign and domestic, may attempt to influence these responses. AI answers should not replace critical thinking at the ballot box.