TIER2 has published a new policy brief titled “Open Science for Artificial Intelligence: Implementing Reproducibility to Promote Trust in AI”. The brief highlights the urgent need for reproducibility to be a central priority in artificial intelligence research.

As AI is becoming ever more central to research methods across domains, including highly sensitive areas such as medicine, many studies remain difficult to reproduce due to cultural, technical, and institutional barriers. This lack of reproducibility threatens scientific integrity, slows scientific progress, and risks undermining public trust in AI and science.

The brief provides actionable recommendations for researchers, funders, publishers, and institutions. These include sharing datasets, code, model cards, and detailed methodologies in line with FAIR principles and community-specific standards. They are supported by investing in infrastructure for reproducible workflows such as open-source machine learning toolkits and trusted repositories, reforming research assessment to reward high-quality and replicable research, and promoting training in reproducibility and Open Science practices.

By implementing these measures, the AI research ecosystem could become more transparent, robust, and trustworthy, ensuring that innovations are credible, reusable, and beneficial to society.