Home > News > “AI, good servant, bad master”

“AI, good servant, bad master”

Published

10 February, 2026

AI is making its way into every aspect of our lives, but how ready are we to adopt it, and under what conditions? A report published in French by the EPFL AI Center, in collaboration with the University of Geneva, presents both the perceptions of a sample of French-speaking population of Switzerland and the recommendations of a citizen’s assembly.

Massive use, despite serious concerns about cyberattacks, deepfakes, and privacy: this is the paradox that emerged from a citizens’ assembly asked to reflect on their attitudes towards artificial intelligence. Initiated by the EPFL AI Center, in collaboration with the Swiss Research Center on Democratic Innovations at the University of Geneva and the Demoscan association, this unprecedented process has resulted in the publication of its final report in French and German. It contains 20 concrete proposals to regulate and support the deployment of AI. In short: AI is a good servant but a bad master.

“AI is one of the most significant technological transformations of our time. Its rapid development affects work, health, education, privacy and democratic life” says Marcel Salathé, co-director of the EPFL AI Center. “Given the scale of these issues, the report reminds us that it is essential for citizens to be able to express themselves and help shape the future of this technology, rather than simply suffering from its effects.”

A two-step approach

This document is the result of two complementary initiatives. First, a survey supported by the Swiss Federal Statistical Office collected the views of 734 residents from the French-speaking regions of Switzerland. The questionnaire focused mainly on the uses and public perceptions of AI.

Among the respondents who expressed interest, 40 citizens were selected to form a diverse panel (canton, age, education level, political interest). The assembly then met over two weekends in November to discuss, debate and deliberate. The process was designed and conducted by the Demoscan association which oversaw the methodology and facilitation, guided by a central principle of neutrality, fostering an informed and balanced discussion.

“Democracy is not limited to the ballot box. A citizens’ assembly is a mechanism that transforms intuitive opinion into reflective judgment: participants have enough time, receive relevant information, work within a neutral framework for debate, and produce argued proposals” says Nenad Stojanović, Professor at the University of Geneva and co-founder of Demoscan.

Widespread adoption but strong expectations

The survey shows that AI has already become mainstream: 87% of respondents have used at least one AI tool, and ChatGPT is by far the most common (70%). But this adoption comes with strong expectations around transparency and regulation. Nine in ten respondents believe systems should clearly indicate when users are interacting with a machine, while 70% want public authorities to strictly regulate AI development.

The most pressing fears relate to malicious use. More than 80% cite hacking and cyberattacks as a major risk, while 77% worry about deepfakes and misinformation, reflecting growing anxiety about the erosion of trust in digital content. Privacy comes next (65%), followed by the impact on jobs (59%). Overall, nearly 69% see AI as a serious threat to privacy and data security.

Who should govern AI?

When asked who should take the lead in governing AI, about a third of respondents point to the Swiss government (31%). Yet nearly as many say they don’t know (27%), highlighting uncertainty around institutional responsibility.

In terms of political priorities, data protection stands out clearly: 68% rank it as the top issue, ahead of ethical guidelines and transparency (41%), and the prevention of uncontrolled AI (40%).

The report is intended to inform academic, institutional, and political debates by grounding them in public expectations. In the preface, participants’ convey a central message: AI systems should not be allowed to make decisions autonomously in ways that weaken individual choice or create dependency.

Salathé hopes the process can now be scaled up. “The next step could be to extend this initiative across Switzerland, so that AI governance is informed by citizen recommendations in all language regions,” he says.

20 proposals, structured around five themes
At the end of the deliberations, the citizens’ assembly produced 20 proposals grouped into five broad areas:

1. Responsible practices: establishing dedicated legislation and educational tools to improve data protection and reduce cyber risks.
2. The role of the State: including the creation of a Federal Office for AI to secure long-term research funding.
3. Access and education: strengthening public awareness of the risks of generative AI and encouraging social interaction to avoid an “all-AI” society.
4. The world of work: preparing for economic and personal impacts of job loss or job transformation, including support for retraining.
5. Traceability: introducing labeling to identify and promote human-made content, alongside stronger copyright protections.

The report is available both in French and German.

This project was supported by the EPFL AI Center and Stiftung Mercator Schweiz.

Author: Melissa Anchisi

Nine ERC Advanced Grants awarded to EPFLresearchers

The European Research Council (ERC) awarded nine “ERC Advanced Grants” to EPFL researchers. This prestigious funding scheme gives senior researchers the opportunity to pursue ambitious, curiosity-driven projects that could lead to major scientific breakthroughs.

(more…)

EPFL researchers create an AI model that thinks like we do

An EPFL team has created a new Large Language Model that is structured similarly to a human brain, allowing users more control and moving away from “black box” AI.

When a standard Large Language Model (LLM) is confronted with a problem, it tries to solve it by matching it to similar information it has seen before, and then give an answer based on those past patterns. But how it decides which information to use and what value it gives to different pieces of information can be somewhat inscrutable from the outside.

The LLM MiCRo (Mixture of Cognitive Reasoners) is architecturally divided into four specialized areas that act like different parts of the human brain, allowing users to have more control over how it approaches a question, and to better understand how it comes to its answers. The model, which was presented at the International Conference on Learning Representations, comes from the NLP Lab, part of the School of Computer and Communication Sciences (IC), and the NeuroAI Lab, part of IC and the School of Life Sciences at EPFL.

The four experts

To create MiCRo, researchers identified four regions of the brains specializing in different functions, which they call ‘experts’: language, logic, social reasoning, and world knowledge.

“The brain is organized into specialized regions, each tuned to handle a specific function. So far, we don’t see this division of labor as clearly in current language models,” says Badr AlKhamissi, a PhD candidate leading this research. “We picked four brain regions that neuroscientists know well and gave the model its own specialized modules, each one trained to be analogous to one of those brain regions.”

An LLM usually functions as a stack of layers that a problem or question can be processed through. In the case of MiCRo, each layer is divided into the four different experts. You give a sentence to the model starting at layer one, for example “The cat is asleep”. Then within this layer, the router can choose one expert for the first word “the”, but a different epxert for second word “cat” and so on, making it modular and highly adaptable.

“Each word of a sentence can go to different experts,” AlKhamissi explains. “So one sentence can actually be processed by multiple experts at each layer.”

Consider a prompt like: “Emma wants to split a CHF 60 dinner bill among three friends, but she knows that Jake lost his job last week and is too proud to say he’s struggling.” A purely mathematical module handles the arithmetic: CHF 60 divided by three is CHF 20 each. But the social reasoning module picks up on something subtler: Emma’s awareness of Jake’s situation, his unspoken pride, and the implicit suggestion that she might quietly cover his share. Both kinds of reasoning are needed to fully understand what’s going on, and in MiCRo, each aspect of the prompt is routed to the expert best equipped to handle it.

“When we see how the model works, we can see that it routes the words that relate to the social aspects to the social expert, and when it does the mathematical part, it routes those numbers to the logic expert.”

This separation makes it easier to see how the model is ‘thinking’ and why it makes certain decisions. It also means decisions can be steered – for example, you can decide to increase the impact of the social expert, or suppress the logic expert, depending on what kind of model you want to use in a certain situation.

“In traditional LLMs, you can do this via prompting by telling the model to make the output more social or make it more related to emotions,” AlKhamissi says. “But here, this is done by intervening in the architecture itself without doing any prompting.”

“A virtuous circle”

To create MiCRo, the EPFL team worked with Greta Tuckute, a neuroscientist from Harvard and MIT, to understand which parts of the human brain are activated by different problems, and then applied that learning to the model.

To identify the region analogous to the ‘logic’ expert in the brain, neuroscientists give humans demanding tasks, such as hard mathematical equations, and less demanding tasks, like easy mathematical equations, and then recorded their brain activity to find which brain regions are the most active for the demanding tasks versus non-demanding tasks. AlKhamissi’s team then did the same for the model, giving it demanding mathematical equations to see which experts would be most activated.

“The cool thing is we just used exactly what they do in neuroscience, but in the model. And the model was able to identify those experts on its own.”

While neuroscience informs the model, the model also informs the understanding of the brain, potentially allowing neuroscientists to discover the contributions of different areas for a given problem or question; for example that a certain sentence activates the language areas 20%, the mathematical areas 50%, and the social reasoning areas 40%.

“For my PhD work, I have been interested in this virtuous circle between neuroscience and AI. In one direction, we use findings and insights from neuroscience about the brain and integrate them into language models,” AlKhamissi says, “and now, with models like MiCRo, we can explore the other direction and ask how we can use AI models to help us understand the brain in a better way.”

Author: Stephanie Parker
Source: EPFL

Smarter waste sorting with AI

EPFL startup WasteFlow has developed an AI-powered copilot that identifies and measures waste streams, helping sorting facilities work more efficiently. Support from several EPFL entrepreneurship programs helped the company accelerate the development of its technology.