Artificial intelligence (AI), especially generative AI systems like ChatGPT, Claude, Perplexity, Gemini, and Mistral Vibe, is increasingly used by students and teachers across all levels of education. This raises important questions: Can AI help schools solve their current challenges, or might it reduce students' ability to think independently and develop real skills, turning teachers into mere assistants? Could AI help reduce educational inequalities, or is it too late to avoid its growing influence on education? These concerns are at the heart of ongoing discussions among educators, students, and policymakers. In response to these concerns, a promising approach has emerged: "educational bilingualism," where AI is treated like a foreign language—something to be learned and used, but not relied on excessively. This idea is being debated widely in the education sector, both in France and internationally. The discussions are complex, as they involve differing views on AI’s role and the fast-paced evolution of the technology itself. AI systems are now embedded at various stages of education, from classroom activities to grading, curriculum development, and even administrative tasks. The central issue is whether generative AI is beneficial or harmful to students and teachers. While AI can support learning by finding relevant resources, explaining complex topics, creating texts and presentations, and offering personalized study plans, it also carries risks. Over-reliance on AI might hinder students from developing critical thinking and problem-solving skills. It could also reduce the role of teachers, potentially leading to issues like academic dishonesty and a decline in the social and interactive aspects of education. For teachers, the situation is similarly complex. While AI can assist in preparing lessons and managing administrative tasks, it may also weaken their ability to design meaningful courses and maintain a critical, pedagogical role. Some educators feel distrust or resistance toward AI, fearing it might undermine the traditional values of teaching. However, the current official approach leaves the decision of how to use AI up to individual teachers, which can lead to inconsistent and unmeasured outcomes. A proposed solution is to implement "digital bilingualism," where education is divided into two tracks: one where AI and digital tools are not used, and another where they are fully integrated. This approach, inspired by bilingual schools, would allow students and teachers to experience both traditional and AI-enhanced learning environments. The "without" track would reinforce foundational knowledge and critical thinking, while the "with" track would help develop digital literacy. This dual approach would help educators and students understand both the benefits and limitations of AI, while protecting against the risks of over-dependence or dehumanization in education. Most teachers are open to using AI, as long as they can choose when and how to use it. However, some are strongly opposed, and public opinion on AI remains divided. Digital bilingualism offers a way for teachers to teach according to their own beliefs—those opposed to AI could focus on the "without" track, while others could incorporate AI in the "with" track. This approach could help balance the potential of AI with the values of traditional education.