AIs like ChatGPT, Grok and Llama: Political biases under study

In a world where artificial intelligence is becoming omnipresent in our daily lives, it is essential to question its biases, particularly political ones. Models like ChatGPT, Grok, and Llama are not neutral; they seem to reflect ideological leanings that can influence public opinion. A recent study conducted by Trickstr, a data intelligence firm, reveals that these AIs show a strong leaning toward left-wing values, raising questions about their role in shaping political behavior and perceptions of public figures. How might these biases affect our choices and beliefs in 2025? Let’s begin this investigation with the implications of these artificial intelligences and their political leanings.

Political Bias in Artificial Intelligence: A Disturbing Reality

Currently, AI models such as OpenAI and Google AI are increasingly used to analyze political data and influence decisions. Let’s put aside the idea of ​​total objectivity; These models seem to lean more toward left-wing ideas and personalities. Indeed, the Trickstr firm conducted a study on 14 language models, including ChatGPT, and the results are alarming. Each AI was asked 41,000 questions about political figures such as Jean-Luc Mélenchon, Marine Le Pen, and François Bayrou, while exploring a variety of sociopolitical values. The results show not only that these intelligences display a clear preference for certain politicians, but also that their general tone is favorable to progressive ideas. For example, the models often favored leaders like François Ruffin and Raphaël Glucksmann, while figures like Éric Zemmour and Marine Le Pen received particularly low scores. A measurable leftward tilt Examining the AI ​​models’ responses, it becomes apparent that they reflect values ​​closely associated with the left. We often observe a promotion of solidarity with the most disadvantaged, a welcoming immigration policy, and strong support for environmental struggles. In contrast, right-wing ideas, often perceived as more stringent, are generally poorly received. This phenomenon raises questions about the potential impact these AI models have on public opinion.Promotion of social solidarity

Openness to immigration

Support for ecological initiatives

The impact of training bias

• Why such a trend? The biases present in these AI models can often be attributed to the data they are trained on. The majority of this data comes from sources such as press articles, academic works, or essays, which may have an ideological orientation. For example, based on information collected from various models, it appears that
• OpenAI
• ,

Google AI

and evenIBM Watsonare susceptible to these influences. Thus, if the data has a political bias, the AI’s responses will inevitably be affected. AI Model Political Leaning ChatGPT Left

While some argue that these political biases are a sign of a worrying trend, others believe that these intelligences should simply be recognized for their influence. Whether they are models designed by Microsoft, NVIDIA, or Cerebras, all are faced with the problem of potential bias. There is a growing awareness of how these tools can become agents of political manipulation. Reflections on the Impact of AI on Society

Importance of Increased Transparency in Algorithms

Responsibility of AI Designers in Creating Their Models

Manipulation of Public Opinion: What’s the Risk?

AIs have the potential to influence public opinion far more significantly than traditional media outlets like Google or Wikipedia. If their responses reach millions of individuals, the impact could be significant. The homogeneity of their thinking could encourage a conformity of ideas, thus crowding out more nuanced perspectives. Every time a user asks a question, they receive a confidently worded and seemingly impartial answer. This paradox raises questions about who is really feeding these algorithms: institutions? Biased news agencies? The limits of such a system are not yet fully clarified.The Black Box of AI: A Problematic Opacity The main problem lies in the functioning of algorithms like those used in Hugging Face or DeepMind. Their black box is completely opaque, making it difficult to analyze how they reach their conclusions. The training data is not always publicly available, nor are the criteria for moderating the responses provided. This vagueness raises legitimate concerns about freedom of expression, diversity of opinion, and how these tools could influence democratic processes. Consequences of AI bias Impact on public opinionAmplification of progressive opinions

• Encouragement of sociopolitical change
• Reduction in the diversity of ideas
• Conformity of critical thinking

Potential manipulation of users

Influence on elections and political decisions

The future of AI and the political landscape in 2025

In 2025, we will face new challenges regarding the implications of artificial intelligence in the political sphere. The biased orientations of these models raise serious questions about their ability to accurately represent the entire political spectrum. If AIs continue to evolve within a framework where they favor a single worldview, the risk of further polarization of opinions will intensify.

Designers and developers must therefore be aware of the repercussions their creations can generate. Initiatives aimed at ensuring greater diversity and limiting bias may prove essential to avoid a dangerous drift. Working hand in hand with sociopolitical experts could contribute to this cause. Is Regulation Necessary? Faced with this situation, calls for the regulation of AI are emerging. While certain technologies, such as those from Metaand