Commission plans whistleblowing mechanism for misuse of AI in science

The European Commission is planning to set up a mechanism that would help researchers report concerns about the misuse of AI in science. This could include establishing an independent EU body or contact point to manage cases of whistleblowing.

This move comes just days after the Commission released a strategy for promoting the development of AI in science.

“Currently, there is no EU-level mechanism to report concerns about the misuse of AI in scientific research,” the Commission writes in a consultation on the future European Research Area (ERA) Act, which opened on October 13. “Researchers lack trusted and secure channels to raise the alarm when AI is used unethically or for (un)intended harmful purposes. This gap increases the risk that dangerous applications go undetected and undermines trust in the research system.”

The Commission also sees a need to align codes of conduct for using AI across the bloc. “Research proposals involving AI are subject to varying requirements on ethics, transparency, intellectual property, data protection and data governance,” it says.

The suggestion is that the ERA Act, which is due to be presented in 2026, would include non-binding EU-wide principles and harmonised guidelines on the responsible and ethical use of AI in research.

The broad proposal has been welcomed by the research community, although questions remain about the approach. 

“Any initiative that helps further clarify uncertainties around the AI use in research represents a step forward,” Rúben Castro, policy and advocacy officer at the Coimbra Group of universities, told Science|Business. However, he questioned the suitability of the ERA Act as a vehicle for the initiative. “Whether this should be dealt with through EU legislation is too early to say,” he said.

In any case, it must be developed with the collaboration of the research community, according to Coimbra director Emmanuelle Gardan. “Universities must be involved in the creation of such mechanism from the outset to ensure that rules are realistic, reflect real research settings, and respect the diversity of research cultures across Europe,” she said.

Populist interference

Julien Chicot, head of research and innovation policy at the Guild of European Research-Intensive Universities, agreed, especially when there is a risk of political interference in research.

“Many of the growing populist parties across Europe are developing an anti-science rhetoric and attacking certain fields of research,” he said. “Against this context [. . .] any whistleblowing mechanism should be better activated by researchers as part of a broader scientific discussion on the quality and integrity of the use of AI in research. Otherwise, the freedom of scientific research could be threatened.”

Related articles

• EU launches AI strategies to boost competitiveness and science
• Zaharieva plans ‘ambitious but realistic’ ERA Act

For Mattias Björnmalm, secretary general of university group Cesaer, addressing the potential misuse of AI in research should not rely solely on compliance mechanisms. “Europe’s approach must be grounded in research integrity – reliability, honesty, respect and accountability – and led by researchers and their institutions,” he said.

Observers have previously said that while AI might be useful to scientists in some circumstances, it is unlikely to be revolutionary. Some even warn that generative AI could create a wave of scientific spam, drowning researchers in low-quality papers and burying genuinely new findings.

Before setting any legislation, both Castro and Gardan believe that the EU needs to review and harmonise institutional, national and EU approaches to the ethical and responsible use of AI in research, in order to avoid duplication or conflicting obligations for scientists. 

“The EU mechanism [. . .] must respect the principle of subsidiarity and therefore complement, enhance, update and link with existing ethics and integrity bodies and mechanisms, not replace them,” Gardan said.