# Generative artificial intelligence in politics: opportunities, challenges and the search for democratic ethics Author: Javier Naranjo Sanjuan Original publication: 2024-10-09 Spanish original: https://estrategiabyaleph.substack.com/p/estrategia-54-ia-generativa-el-motor English URL: https://elcontemplador.github.io/estrategia-english/essays/054/ Status: Published translation This is a translation of the original Spanish essay published on 9 October 2024. Its claims, examples and forecasts retain that historical context. English publication: 2026-09-29 *Historical note: this article and its screenshots record a 2024 experiment. Product names, technical descriptions and the model’s responses are preserved in their original context.* > **This week’s main article in estrategIA is by [Javier Naranjo Sanjuan](https://www.linkedin.com/in/javiernaranjosanjuan/)**, an IT specialist in intelligent systems and a public policy consultant working on innovation and entrepreneurship. **As a new departure**—and this is an area in which we would like to bring you interesting developments in future—**he has also generated it as the video below.** *Archive note: the preserved local source announces a video here but contains no playable video or video link. The recording has not been reconstructed.* We live amid constant revolution. A few years ago it was blockchain technology; now artificial intelligence is advancing in leaps and bounds, transforming every field there is. When I was asked to write this article, I posed a fundamental question: how many types of generative artificial intelligence are there? There are various types of generative AI, each with its own techniques and applications. Generative language models such as ChatGPT (from OpenAI), Bard (from Google) and Bing Chat (from Microsoft), for example, are designed to generate coherent text. Other types include Generative Adversarial Networks (GANs), which create images and videos through competition between two neural networks; autoregressive models, which predict sequential data; and Variational Autoencoders (VAEs), which generate new data samples. There are also Transformers, effective at generating both text and images; diffusion models, which produce realistic images from random noise; and generative reinforcement learning models, used in games and simulations. Multimodal generative AI can handle different types of content; recurrent neural networks (RNNs) and LSTMs are useful for analysing long-term dependencies; and there are specific models for audio and music, such as WaveNet and Jukebox. Finally, grammar-based models can create complex structures while respecting syntactic rules. Although both AI assistants and custom GPT models use natural language processing technology, there are key differences in their purpose and degree of specialisation. Assistants such as ChatGPT are designed for general interaction with users, offering broad responses adaptable to a variety of subjects and contexts, with no particular focus beyond being useful and safe in everyday conversations. Custom GPTs, by contrast, are models adapted to particular tasks and audiences; they are trained on specific data to address particular subjects with greater precision and depth. In politics, for instance, they can generate content aligned with a particular ideological or communications approach. This customisation allows GPTs to be more effective in specialist settings, while assistants aim to be versatile and useful to a wide range of users. ## A custom GPT for estrategIA Generative AI in politics can also automate tasks that traditionally required considerable human involvement, such as writing speeches, preparing responses for debates or generating promotional materials. With AI’s help, these activities can become faster and cheaper. Yet this same automation can create a disconnect with voters, who may perceive excessive use of technology as a lack of authenticity or an attempt at manipulation, rather than a sincere effort to connect. Here is an example of an experimental custom GPT using the estrategIA articles written to date. [![The 2024 custom GPT configuration for estrategIA, with description, instructions, conversation starter and uploaded PDF issues.](https://elcontemplador.github.io/estrategia-english/assets/images/aa95eb5b-4678-4741-94e4-857d7308fbc2_989x565.png)](https://substackcdn.com/image/fetch/$s_!-kz8!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Faa95eb5b-4678-4741-94e4-857d7308fbc2_989x565.png) *Image 1. Uploading the custom knowledge base.* **Accessible translation of the configuration shown:** The GPT is named “estrategIA”. Its description and instructions both read “Artificial intelligence in politics and government”. The conversation starter is “Hello estrategIA”. In the “Knowledge” section, PDF files of previous issues are attached; their filenames are partly truncated. The interface explains: “If you upload files to Knowledge, conversations with your GPT may include the contents of the files. Files can be downloaded when Code Interpreter is enabled.” The preview repeats the GPT’s name, description and greeting. The configuration is marked active and visible only to the owner, with updates pending. [![A custom GPT summarises six themes in the estrategIA archive and drafts a tweet; the full conversation is translated below.](https://elcontemplador.github.io/estrategia-english/assets/images/8ef8d328-d3ec-4c01-ba10-f283bbd286b6_1008x1468.png)](https://substackcdn.com/image/fetch/$s_!jopG!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8ef8d328-d3ec-4c01-ba10-f283bbd286b6_1008x1468.png) *Image 2. Example of a customised conversation.* ### Accessible translation of the conversation *The following exchanges reproduce the historical model output in the screenshot. They are translated examples of what the GPT said, including its characterisation of the archive.* > **User:** Hello estrategIA, what do you have in your knowledge base? Can you give me a summary? > > **GPT:** In my knowledge base I have a series of articles and documents focused on the intersection of artificial intelligence (AI) with politics and government, exploring both current use and the possible futures of AI in these fields. Here is a summary of some prominent themes: > > 1. **Four golden rules for working with AI according to Ethan Mollick:** Mollick, a professor at the Wharton School, proposes strategies for integrating AI effectively into everyday work. The rules include always inviting AI to participate, maintaining human oversight, customising the AI’s “personality” to suit the task, and assuming that the technology will continue to evolve. > 2. **Stories of politics and AI in 2045:** A series of fictional stories exploring future scenarios in which AI plays a central role in politics, such as an AI assistant becoming an independent presidential candidate, or the creation of political truthfulness laws requiring AI to intervene in political debates to ensure honesty. > 3. **Innovations in political data analysis with AI:** Using advanced models such as Google Gemini Pro 1.5 and GPT-4o to improve political data analysis and the comparison of election manifestos, providing quick summaries and comparisons of proposals from multiple parties, and detailed analyses of turnout data. > 4. **AI’s impact and controversies on social media during elections:** An analysis of AI’s impact on elections, focusing on its use on social media to influence public opinion and on related controversies, such as data manipulation and disinformation. > 5. **Integrating AI into public administration and defence:** Examples such as the Community of Madrid’s “Cuenta Digital” application, which uses AI to simplify public procedures, and the development of AI assistants by Spain’s Ministry of Defence to improve operations and security. > 6. **Challenges and opportunities for AI in the European regulatory environment:** A discussion of how the European Union’s strict regulations are creating a “two-speed AI”, with faster advances in the United States and China than in Europe because of regulatory restrictions. > > These are just a few examples of the topics covered in the documents in my knowledge base. If you are interested in a particular topic, I can explore it further or look for more details. > > **User:** Write me a summary tweet. > > **GPT:** 🌐💡 Artificial intelligence is transforming politics: from improving data analysis and comparing election manifestos to imagining futures where AIs are candidates. Innovation, ethical challenges and opportunities are intertwined in this new era. 🚀🤖 #PoliticsAI #DigitalFuture #TechnologicalInnovation ## Transparency, trust and democratic responsibility There is also a growing debate about transparency in the use of these tools. Should campaign teams disclose when and how they use artificial intelligence to generate content? This question affects not only the ethics of political campaigns, but also public perceptions of the legitimacy of their communications. In a world increasingly aware of AI’s role in society, transparency may become a key element in maintaining public trust. Despite the challenges, generative AI offers enormous possibilities, and its impact will depend largely on how these tools are used. At its best, generative AI could democratise access to information, enabling more people to understand complex political issues through automatically generated explanations that are clear and accessible. It could also facilitate better-informed debates and more active civic participation. We must remember, however, that while generative AI can be a powerful tool, it cannot replace the importance of human dialogue, empathy and real connections between political leaders and their communities. Technology can help amplify messages, but true connection is built through tangible actions and a sincere commitment to listening and responding to people’s needs. Ultimately, the responsible and ethical use of generative artificial intelligence in politics could make the difference between a future in which technology empowers people and promotes democracy, and one in which it is used to manipulate and polarise. The key therefore lies in how we choose to use the technology. Well-managed AI can be a valuable tool for improving communication and effectiveness, but it must never replace integrity, transparency and an authentic commitment to democratic values. After writing the text, I wondered: would this article be in line with previous estrategIA articles? So why not ask it? [![The custom GPT assesses the article against four themes in its knowledge base; the full response is translated below.](https://elcontemplador.github.io/estrategia-english/assets/images/50f8d134-c6c6-493b-9499-8a216b470d20_844x756.png)](https://substackcdn.com/image/fetch/$s_!daUG!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F50f8d134-c6c6-493b-9499-8a216b470d20_844x756.png) ### Accessible translation of the GPT’s assessment *Historical model response shown in the screenshot:* > Yes, the text you provided is in line with the ideas present in the articles in my knowledge base. It reflects several key themes also addressed in the documents I have available: > > 1. **The transformation of politics through AI:** The text mentions how artificial intelligence, especially generative AI, is changing the political landscape. This is a recurring theme in articles about automating tasks, creating content, and the new opportunities and challenges this presents for political actors. > 2. **Different types of generative AI and their applications:** As explored in the articles, the text details the different types of generative AI, from language models such as ChatGPT to more specific models such as Generative Adversarial Networks (GANs) and Transformers. This is consistent with the discussions of AI’s various applications and technological capabilities in politics and beyond. > 3. **The impact and ethics of using AI in politics:** The text addresses transparency and authenticity in the use of AI tools for political campaigns. This aligns with the ethical concerns mentioned in the documents about AI’s use in political decision-making and communication with the electorate. > 4. **Opportunities and challenges for AI in politics:** It also explores both the opportunities and the potential dangers of using generative AI, reflecting the balanced perspective of articles that discuss how AI can democratise access to information but may also lead to risks of manipulation and disconnection from voters. > > In summary, your text is in tune with the ideas expressed in the articles in my knowledge base, addressing both the advantages and the concerns about applying AI in politics and the importance of using this technology ethically and responsibly. As we explore the boundaries of what is possible with these new technologies, we must not forget that human beings remain the true drivers of change and progress. Artificial intelligence can help shape discourse, but the decisions that affect our lives and societies must still rest with real people who are accountable and committed to the common good. Technology is only a tool, and true intelligence lies in knowing how, when and why to use it. [Javier Naranjo Sanjuan](https://www.linkedin.com/in/javiernaranjosanjuan) *IT specialist in intelligent systems and public policy consultant on innovation and entrepreneurship, [@javiernsanjuan](https://x.com/JavierNSanjuan)*