Pause IA is a French non-profit run by volunteers that warns citizens and public authorities about the dangers of an artificial intelligence developed too fast, without safeguards. We call for a global, coordinated pause on the training of the most powerful AI systems, until we know how to make them safe and democratically controlled. We are the French branch of the international movement PauseAI, present in around fifteen countries.

Pause IA brings together a leadership team of about ten people, led by Maxime Fournes (president and co-founder) and Clémence Peyrot (executive director), along with more than a hundred volunteers. Our backgrounds are varied: engineers, teachers, researchers, lawyers, psychologists, students, retirees. What unites us is a shared concern about the current trajectory of AI, and the conviction that citizens have a say.

Learn more: Who we are.

Pause IA runs on donations from individuals and a few grants from independent foundations committed to AI safety. We receive no funding from artificial intelligence companies, their investors, or the states racing to build AI. Most of our work relies on volunteers.

No. Pause IA is independent of any political party, company or state. The risks of poorly controlled AI concern everyone, on the right as on the left, in France as elsewhere. We work with anyone willing to take these risks seriously: elected officials from different sides, scientists, unions, associations. We do not align with any partisan or commercial agenda, and our positions are public.

Yes. Pause IA is the French branch of an international movement, PauseAI, born in the Netherlands in May 2023. In less than three years, local groups have formed in around fifteen countries across four continents: the United States, Canada, the United Kingdom, Germany, Spain, Italy, Sweden, Australia, Nigeria, Kenya, and others. They all share the same demand: a global pause on the training of the most powerful AI systems, as long as we do not know how to make them safe.

No, quite the opposite. We believe that real innovation is controlled innovation, one that serves humanity instead of putting it at risk. No one would accept a medicine sold without testing, or a plane flying without certification: yet that is what AI labs do today. Asking for safeguards is not holding back innovation, it is steering it toward what helps us progress rather than what endangers us.

Today's AI is already used to spread disinformation at scale, to influence elections, to carry out cyberattacks and to weaken entire sectors of employment. The future is even more worrying: the labs of the tech giants are building ever more powerful systems that they themselves admit they cannot control. The risk goes as far as biological attacks made easier by AI, or even a loss of control that could threaten humanity as a whole.

Learn more: The dangers of AI.

AI can be of great service in medicine, for the climate or in research, we do not deny that. These benefits can come from specialised AI: tools designed for a specific task, limited to their domain and, for that reason, easier to control. Yet that is not the direction taken today by the billion-dollar companies that dominate this industry. Their goal is to build general-purpose AI, capable of replacing humans in a growing number of tasks, in order to capture the economic benefits and gain a geopolitical advantage.

A lot of damage is already here: mass disinformation, cyberattacks, breaches of privacy, psychological and cognitive strain, the first job losses. The most serious dangers, such as engineered pandemics or the loss of control of autonomous systems, could become real within a few years. The leaders of the big AI companies themselves are announcing human-level artificial intelligence within a few years.

Yes, and for a long time. In 2023, more than 600 researchers and AI executives, including the heads of the three main labs (OpenAI, Google DeepMind and Anthropic), signed a statement affirming that the risk of extinction from AI should be a global priority, alongside pandemics and nuclear war. Since then, the warnings have only multiplied. In September 2025, the global call to set "red lines" for AI, launched at the UN, brought together more than 300 leading figures, including former heads of state, Nobel laureates and AI pioneers awarded the Turing Prize. And in 2026, the International AI Safety Report, chaired by Yoshua Bengio (Turing Prize) and modelled on the IPCC, confirmed these dangers. These are not fringe fears: it is a scientific consensus that broadens year after year.

A moratorium is a temporary suspension of an activity, usually decided by an authority (a government, an international organisation). Concretely, we ask that labs stop training their most dangerous frontier general-purpose AI models, not that they shut down. This pause would last as long as needed to put in place the rules, controls and technical tools that would allow work to resume safely. Similar moratoriums already exist, on ethical issues (cloning), technological ones (autonomous weapons, GMOs) or environmental ones (deep-sea mining): they are often temporary, while legal frameworks are built or more thorough scientific assessments are carried out.

Learn more: What we ask for.

No. Our demand concerns the training of the next generations of AI, far more powerful than the ones you use today. ChatGPT, automatic translators, writing assistants or medical imaging AI are not our target. What worries us are the systems currently in development, designed to match or exceed human intelligence across all topics, and that no one today knows how to control.

Humanity has already managed to coordinate to establish global moratoriums, notably on biological weapons, human cloning or nuclear testing. And in practice, Pause IA is not starting from scratch: the UN, the G7 and UNESCO have already set out principles and codes of conduct on AI, while experts have drawn up concrete proposals to govern these systems. Moreover, the idea of a moratorium is far from marginal: in France (CeSIA / OpinionWay survey) as in the United States (Future of Life Institute survey), a large majority of the population wants to slow down or better regulate the development of AI, while only a tiny minority supports the current pace.

There are many ways to help, depending on the time you have. You can join the association, become a volunteer, join a local group near you, sign our petitions, write to your representatives, donate, or simply talk about the topic around you. Everything counts.

To discover all the possible actions: Take action.

Absolutely not. Our volunteers are teachers, artists, students, parents, engineers, retirees… What matters is taking the topic seriously and wanting to contribute in your own way. We need very varied profiles: communication, writing, events, legal, translation, welcoming newcomers, graphic design, video, organisation. You can help.

Becoming a member involves a modest fee, which helps fund our actions and our independence. If it is an obstacle for you, talk to us: we will find a solution. You can also take part actively as a volunteer, join a local group or follow our campaigns without being a member.