Artificial intelligence is moving into a territory that is far more sensitive than simple automation. Chatbots are no longer used only to write emails, translate documents, summarize reports or generate ideas. Millions of people now ask them about wars, elections, political parties, governments, controversial historical events and even which political position may be closest to their own.
That changes the role of the technology.
A search engine mainly points users toward information. A chatbot can instead select information, organize it, interpret it and return a single conversational answer. The user may therefore see not only the information itself, but an AI-generated interpretation of that information.
That raises a question that will become increasingly important: who decides what an AI chatbot tells us?
The answer is not as simple as saying that governments control ChatGPT, Gemini, Claude or Grok. There is no evidence that a government can simply dictate every answer produced by a major commercial AI model. But governments can influence the environment in which AI systems operate through legislation, regulation, restrictions, public-sector deployments and national rules.
At the same time, technology companies influence their models through training data, system instructions, safety policies, fine-tuning, moderation systems and the sources their products are allowed to access.
And then there is the user.
The wording of a question can change the answer. The model can change the answer. The available sources can change the answer. The country in which the system operates can matter. Even the language used to ask the question can alter the information environment.
The result is a complicated chain of influence rather than a single political hand controlling the machine.

The first layer: governments can regulate AI without writing its answers

The most obvious mechanism is legislation.
The European Union is one of the clearest examples. The AI Act does not tell ChatGPT what political opinion it should express. It does not instruct an AI model to support one party or reject another.
Instead, it establishes boundaries around how certain AI systems can be developed and used.
European rules prohibit certain manipulative or deceptive AI practices when they are designed or have the effect of materially distorting a person's behaviour and causing, or being reasonably likely to cause, significant harm. The European framework also includes transparency requirements for certain AI-generated content and systems that interact directly with people.
Several important AI Act provisions became enforceable on 2 August 2026.
This distinction matters because regulation and political control are not the same thing.
A government can say, for example, that an AI system must inform users that they are interacting with a machine, or that certain manipulative practices are prohibited. That does not mean the government has written the model's political opinions.
But regulation still shapes the boundaries within which the technology operates.

The hidden influence of company rules

The second layer is less visible to ordinary users: the decisions made by AI companies themselves.
Large language models are trained using enormous collections of material. The composition of those datasets matters. So does the filtering process. So do the instructions used during training and the human feedback used to determine which answers are considered better.
Then come the guardrails.
These systems are designed to prevent harmful, illegal, dangerous or abusive outputs. That is an essential part of building a responsible product.
But every guardrail involves a decision.
Someone has to determine what counts as harmful, what should be refused, how uncertainty should be expressed and when a political question requires additional context.
That does not mean engineers are secretly programming an ideology into the model. It means that human choices inevitably become part of the behaviour of systems created by humans.

Training data is not politically neutral

There is another problem that is easy to overlook.
The internet is not a perfectly balanced representation of humanity.
It contains academic research, government documents, professional journalism, social media, political propaganda, personal opinions, misinformation, entertainment, corporate material and countless other forms of communication.
A model trained on large quantities of internet material inevitably learns patterns from that environment.
Researchers have pointed out that much of the information shaping major AI systems reflects the cultural characteristics of the societies that produce the largest share of digital content.
That does not automatically create a left-wing or right-wing chatbot.
It does mean that the idea of a perfectly neutral machine learning from a perfectly neutral world is unrealistic.

Are AI chatbots politically biased?

This became one of the most controversial AI questions of 2026.
A major Washington Post investigation tested several AI systems using political questions based on previous academic research. The results showed substantial differences between models.
In those tests, OpenAI's model produced left-leaning arguments far more often than right-leaning ones, while Google's Gemini much more frequently presented both sides of a political question. Other systems produced different combinations.
Those findings attracted enormous attention.
But they should not be turned into the simplistic statement that “AI is left-wing”.
A language model does not have a political identity in the same way a human voter does. It generates responses according to its training, instructions, context and available information.
A different prompt can produce a different response.
A different model can produce a different response.
A different source environment can produce another answer.
The experiments therefore tell us something important about model behaviour, but they do not prove that a government has secretly ordered a chatbot to support a particular ideology.

The Japanese election experiment

A particularly interesting development came from research involving Japan's February 2026 general election.
Stanford researchers created tens of thousands of synthetic voter profiles with different political positions, demographic characteristics and geographical backgrounds. They then asked several AI models to recommend political parties.
The results showed that political recommendations could change depending on the ideological profile of the hypothetical voter.
But the study highlighted something even more important: the information environment available to the AI mattered.
This means that the political influence of a chatbot may not come from a single hidden instruction.
It may emerge from the interaction between the model, the sources it can access and the way a political question is framed.
That is a much more complicated problem.

America.gov offers a real-world example

At the end of September 2026, the debate became much more concrete in the United States.
The Trump administration launched a redesigned America.gov, a government portal intended to make federal information and services easier to access. According to the General Services Administration, the platform combines information from more than 29,000 government websites with AI-enhanced search technology.
The idea is straightforward: instead of forcing citizens to navigate thousands of separate government pages, AI can help them find the relevant information.
But politically sensitive questions immediately exposed the difficulty.
During early testing, the system produced answers that contradicted some claims made publicly by President Donald Trump, including statements about the 2020 presidential election.
Later, the system began refusing some political questions altogether, saying it could not answer political questions.
That episode does not prove that the government ordered the AI to lie.
It does demonstrate something more fundamental: the configuration of an AI system can change what users see.
Change the sources. Change the instructions. Change the guardrails. Change the response policy.
The experience can change.

What happens when government becomes the information source?

This creates a difficult question.
Using government sources is not automatically propaganda. In fact, for administrative information, official sources are often exactly what an AI system should use.
If someone asks how to renew a passport, apply for a government benefit or understand a tax form, official documentation is generally preferable to an anonymous social-media post.
The situation becomes more complicated when the subject is political history, elections, war, foreign policy or controversial government conduct.
Then the selection of sources becomes extremely important.
If the AI sees only one side of an issue, the answer will inevitably be shaped by that information environment.
And this principle applies equally to governments and private companies.

Censorship does not always look like censorship

In authoritarian systems, the relationship between politics and AI can be much more direct.
Governments can establish legal boundaries around what may be discussed. Sensitive political subjects may be restricted. Criticism of political leaders may be limited. Certain historical events may be described differently from how they are presented elsewhere.
Recent research has found differences in how AI systems respond to questions concerning political regimes and restrictions on free expression.
Again, this does not mean every AI model is censored in the same way.
It means that the jurisdiction in which an AI system operates can influence the information environment surrounding it.
For global AI companies, this creates a difficult balancing act between different national laws, corporate policies and expectations of users.

The biggest danger may be trust

The most important issue may not actually be political bias.
It may be trust.
People know that politicians want to persuade them. They know that political parties have interests. They know that newspapers have editorial lines and that social-media influencers compete for attention.
A chatbot looks different.
It speaks calmly. It organizes information. It can compare arguments. It can summarize hundreds of pages in seconds.
That can create the impression that the machine is an impartial referee.
But it is not an oracle.
A chatbot can misunderstand a question. It can rely on incomplete sources. It can make factual mistakes. It can omit relevant information. It can respond differently depending on the wording of a prompt.
And it can change over time as its model, instructions, sources or safety systems change.

Can AI influence voters?

There is already evidence that AI-generated political messages can be persuasive.
Research from Stanford has found that AI-generated political arguments can influence opinions at levels comparable to arguments produced by humans under certain experimental conditions. Other research suggests that people may be more receptive to an argument when they believe it comes from an AI system because they perceive the technology as less politically motivated.
That creates an unusual paradox.
The more people believe an AI is neutral, the more powerful its influence could become.
A chatbot does not need to tell someone “vote for candidate X”.
It may influence the voter simply by explaining why one candidate appears more consistent, why another has a controversial record, why one policy is economically stronger or why a particular political position appears more reasonable.
The selection and framing of information can matter as much as an explicit recommendation.

Italy is entering the same territory

Italy is approaching an important transition.
The 2022 general election took place before ChatGPT became publicly available. The political environment of the next national election will be very different.
Italian voters can already ask AI systems to compare political programmes, explain policy proposals, summarize parliamentary records, examine statements by politicians or identify which party appears closest to a particular set of views.
That can be extremely useful.
But it creates a new responsibility for users.
An AI-generated summary should not replace the original sources.
A chatbot can help a voter navigate information. It should not become the only information source.
The difference is crucial.
If an AI system becomes the only intermediary between citizens and political information, whoever determines its information environment gains enormous influence.

The future will require transparency

The answer is probably not to ban AI from political discussion.
That would be both impractical and potentially harmful.
A better approach is transparency.
Users should be able to understand where important information comes from. They should know when an answer depends on live web retrieval, when a claim is uncertain and when the system has been instructed to refuse or restrict a particular type of content.
AI systems should also make it easier to compare sources rather than presenting a single apparently definitive answer.
A political chatbot that says “this is the truth” is far less useful than one that can explain which facts are well established, which interpretations are disputed and which claims come from competing political actors.

Absolute neutrality may be impossible

The final lesson may be the simplest.
An AI system cannot be completely separated from human choices.
Someone chooses the training material. Someone designs the model. Someone defines the safety rules. Someone decides which sources are available. Someone determines how the system should respond to sensitive questions.
Even the decision to be “neutral” involves choices about what neutrality means.
The real goal, therefore, may not be an impossible perfectly neutral machine.
It may be an AI system that is transparent enough for users to understand its limits, sources and uncertainty.
Because the political battle of the future may not be fought only through television, newspapers, social networks and campaign platforms.
It may also be fought through the invisible layer between a citizen's question and the answer appearing on a screen.
And the most important political question about AI may ultimately be this:
Who decides what the machine gets to tell us?

🔗 RELATED ARTICLES

Fideuram Scam: €95 Million Transferred Using AI — Cloned Voice, Fake Messina and €39.5 Million Still Missing
https://www.commentalanotizia.com/2026/09/truffa-fideuram-intelligenza-artificiale-95-milioni.html
OpenAI’s AI Agent Bypasses Australian Medicare Restrictions: What Really Happened
https://www.commentalanotizia.com/2026/09/agente-ia-openai-medicare-australia.html
Revolut Case: The Fake PEC Scam and the Hackers’ Ransom Demand — What Really Happened
https://www.commentalanotizia.com/2026/09/revolut-attacco-hacker-furto-dati-pec.html
Instinct: The AI That Can See Your Screen — The New Frontier of Privacy Is Already Here
https://www.commentalanotizia.com/2026/09/instinct-ai-privacy-schermo-dati-personali.html
Facial Recognition in Italy: What Is Really Changing
https://www.commentalanotizia.com/2026/08/riconoscimento-facciale-italia-ai-act.html
OpenAI’s AI Agent That Hit Hugging Face: What Really Happened?
https://www.commentalanotizia.com/2026/07/openai-agente-ia-hugging-face-incidente.html
Google Wallet Tracks Online Purchases: What You Need to Know
https://www.commentalanotizia.com/2026/07/google-wallet-traccia-acquisti-online.html
Italy Approves AI Decrees: What Really Changes
https://www.commentalanotizia.com/2026/06/italia-decreti-ai-intelligenza-artificiale-lavoro-scuola-sicurezza.html
Artificial Intelligence and the Future of Work: What Will Really Change?
https://www.commentalanotizia.com/2026/03/intelligenza-artificiale-futuro-lavoro-ai.html

👉 Want to read the full original article? Read it on Commenta La Notizia:👇
https://www.commentalanotizia.com/2026/10/intelligenza-artificiale-politica-governi-chatbot.html
The original article is in Italian. You can use your browser’s automatic translation feature if needed.

Subscribe and vote for all articles 👇
https://www.publish0x.com/goldweb?a=xkazKWYYbJ

💬 Join us on Telegram to comment and follow all the insights
👉 https://t.me/CommentaLaNotizia

© 2026 CommentalaNotizia.com - All rights reserved
Official sources used under fair use

📺 Follow Commenta La Notizia for deeper insights, global analysis, and fearless discussions about the issues that shape our world — from politics and technology to society and the hidden forces behind the headlines.
⚠️ Legal Notice This article and the related multimedia content are the exclusive property of the author. Copying, reproduction, distribution, or modification in any form or on any platform is prohibited without written permission.
Copyright © CondividiSulWeb – YouTube Commenta La Notizia

👤 Author: @condividisulweb

✍️ Commenta La Notizia is my editorial project created to provide clear and engaging insights into news, events, and major current issues.
Every article, video, and short is made with a clear goal: not only to inform, but also to give context, reflection points, and analysis to understand not just what happens, but especially why it happens.
With Commenta La Notizia I want to build a reliable and accessible space where everyone can get informed, understand, and join the discussion.
🌐 If you want to stay updated, you can follow me here.
💬 Your opinion matters! Every piece of news is a chance to discuss: share your thoughts and make your voice heard.
📰 Comment, share, stay informed.

🙏 Thank you for reading: your time and your opinion are the true value
https://www.commentalanotizia.com/

✍️ SUPPORT FREE INFORMATION✍️

⚠️REMEMBER, YOU CAN VOTE ON ALL ARTICLES EVERY HOUR
AND REGISTER FOR FREE TO VOTE!⚠️

👇👇👇👇 😍 HELP GROW MY PASSION😍 👇👇👇👇

• CommentaLaNotizia.com

• Telegram

• Publish0x

• Blurt

• Steemit

• Facebook

• X

• YouTube

• Pinterest

• Other platforms coming soon