European Commission updates AI guidelines for research

The European Commission has released the fourth version of its ERA Living Guidelines, providing updated frameworks for the responsible use of generative AI in research. This update introduces specific guidance on agentic AI—systems capable of pursuing complex goals with varying degrees of autonomy.

The guidelines emphasize that researchers maintain full accountability for their work, regardless of the level of assistance provided by AI agents. Using such technology cannot be viewed as a delegation of responsibility. Consequently, researchers are instructed to implement human checkpoints to validate methodological choices, hypotheses, data processing, and conclusions. The document explicitly states that AI systems cannot be credited as authors, co-authors, or owners of research.

Regarding transparency, the guidelines establish new standards for disclosure. Researchers must disclose the substantial use of AI tools for tasks such as literature reviews, hypothesis development, or data interpretation. However, minor applications like grammar correction or formatting do not require disclosure. The update also addresses concerns regarding reproducibility, noting that the variable nature of generative AI outputs may impact the consistency of research results.