The French National Center for Scientific Research (CNRS) has raised the alarm over a growing crisis in the scientific publishing world, driven by the rise of artificial intelligence (AI) and the pressures of a profit-oriented publishing model. Researchers are increasingly using AI tools to generate entire articles, which has led to a flood of questionable content online. This trend threatens the reliability of scientific knowledge, as AI-generated work is becoming harder to distinguish from legitimate research. The CNRS is responding by developing new detection tools and overhauling how it evaluates its scientists, aiming to restore trust in research integrity. The crisis came to light two years ago when a respected journal published an article illustrated with images of a rat with exaggerated and fantastical features, all created by the AI tool Midjourney. This incident exposed a major flaw in the peer review process, which is meant to scrutinize every detail of a study before publication. Today, AI tools are being used not only to generate content but also to manipulate the peer review system. Some researchers are hiding instructions that tell AI how to produce results, making it harder for reviewers to detect fraud. These practices have been uncovered on platforms like Arxiv, with involvement from researchers at top institutions. At the heart of the issue is the "author-payer" model, which shifts the cost of publishing from readers to research institutions. Under this model, labs pay high fees to publish their work, creating a strong financial incentive for journals to accept more articles. This has led to a decline in the quality of peer review, as some journals prioritize quantity over rigor. In France, this system has already cost millions in research funding, and the costs could soar to 200 million euros annually if left unchecked. The consequences go beyond money—scientific progress relies on a foundation of reliable research, and false studies can destabilize the entire field. To combat this, the CNRS is deploying tools like BibCheck, which scans research manuscripts for references to discredited or AI-generated studies before they are published. At the same time, the organization is rethinking how it evaluates scientists, moving away from a system that rewards sheer volume of publications. Instead, researchers are now assessed based on the diverse impact of their work, including public engagement and leadership. These changes are part of a broader international effort, led by the CoARA coalition, to promote a more ethical and transparent approach to scientific research. As the CNRS and others work to restore integrity, the challenge remains to ensure that global research practices follow suit.