MEPs call for human control over AI and international rules to protect democracy

The European Parliament’s Committee on Constitutional Affairs calls for human oversight of the use of artificial intelligence by EU institutions, systematic assessment of risks to elections, and an international agreement on the major risks of AI. The report also proposes an AI observatory in Parliament and new protections related to digital identity, algorithmic transparency, and discrimination.

European Union institutions should retain human control over the use of artificial intelligence, and the European Parliament should systematically assess the effects of AI on elections, disinformation, and the functioning of democracy, according to a report adopted by the Committee on Constitutional Affairs. The document goes beyond technical rules for AI systems and also proposes an international agreement on security and societal risks, new fundamental protections, and the creation of an AI observatory within Parliament.

In brief

AFCO adopted the report with 21 votes in favour, three against, and no abstentions. The plenary is expected to vote on it during the 19–22 October session.

The use of AI by EU institutions should remain people-centred, subject to human oversight, and compatible with data protection and the Union’s values.

MEPs call for systematic assessments of the effects of AI on the integrity of elections, disinformation, manipulation, foreign interference, privacy, and the use of deepfakes.

The report proposes an international agreement on the major risks of artificial intelligence, inspired as a model of cooperation by international nuclear non-proliferation regimes. However, this vote does not mean that international negotiations have already been launched.

AFCO also proposes a Parliament AI observatory, with no additional budget and no regulatory powers, as well as the use of AI to analyse legislation, with a ban on replacing human political responsibility.

The report, prepared by MEP Manolis Kefalogiannis of the EPP, Greece, is based on the idea that the development of artificial intelligence has also become a constitutional issue. The use of AI systems can affect how public decisions are made, how citizens receive information, and even the relationship between institutions, administration, and democratic oversight.

AFCO calls on European institutions to apply themselves the standards that the Union promotes for the responsible use of artificial intelligence. Systems used in the European administration should remain under human oversight and comply with the protection of personal data and the other fundamental values of the EU.

At the same time, the report advocates giving priority to European AI tools in the Union’s institutions and for coordinated investment to enable the development of its own technological capabilities. The objective is linked to what MEPs describe as strategic digital autonomy.

This approach does not mean eliminating technologies developed outside the EU as a result of the report. The document sets out a political direction for the use and procurement of technology by European institutions, not a general ban on AI systems from outside the Union.

An important part of the report concerns electoral democracy. AFCO calls on Parliament to systematically assess both the risks and the opportunities created by AI for democratic processes.

The risks listed include disinformation, information manipulation, foreign interference, violations of privacy, and the use of synthetic content to influence voters. Deepfakes are treated as an example of how technology can make it more difficult to distinguish between authentic material and artificially generated content.

However, the report does not present AI exclusively as a threat. MEPs believe that systems used transparently and with accountability mechanisms can improve citizen participation and help administrations and legislators manage large volumes of information.

AFCO considers that the current European electoral framework is increasingly difficult to apply to risks that are developing alongside artificial intelligence. This is one of the reasons why the report addresses the issue beyond the current AI Act.

The European regulation on artificial intelligence is already the general framework for placing AI systems on the market and using them in the Union. The AFCO report does not replace it and does not itself amend its rules. It focuses on institutional and democratic aspects that the parliamentary committee considers insufficiently covered by the current framework.

One of the most extensive proposals is to initiate a discussion on new protections in the EU Charter of Fundamental Rights. The document mentions protection of personal identity and image, algorithmic transparency, digital self-determination, protection against algorithmic discrimination, and the principle of liability for the harm caused.

These formulations do not represent new rights already introduced into the Charter. The AFCO vote expresses a political request to amend or supplement the existing framework. The Charter remains in its current legal form until any separate amendment procedure is undertaken.

The international dimension is one of the report’s most unusual components. MEPs call on the EU to promote an international agreement for AI technologies that may generate major security risks or societal effects.

The parliamentary committee uses the nuclear non-proliferation regime as inspiration for the idea of an international control architecture. The comparison concerns the need for cooperation between states to manage a technology with cross-border effects and does not equate the risks of artificial intelligence with those of nuclear weapons.

The report does not announce the existence of a treaty already under negotiation, nor does it mandate the European Commission to sign such an agreement. It is a political request addressed to the EU to try to play a leading role in building global rules.

AFCO simultaneously calls for any potential safeguards not to unnecessarily hinder innovation or competitiveness. The document thus seeks to combine the demand for rules and democratic accountability with concern for the ability of European companies to develop competitive technology.

At Parliament level, the report proposes creating its own AI observatory. The structure would use expertise already available and operate without additional budgetary resources.

The observatory would not be a regulatory authority and would not receive powers to authorise or sanction AI systems. Its role would be to support the work of parliamentary committees.

Suggested tasks include post-adoption assessment of legislation, monitoring the practices of European institutions, and analysing the effects of AI on the institutional balance and democratic processes.

The observatory could also identify overlaps, gaps, or contradictions between the various European acts that directly or indirectly regulate the use of artificial intelligence. Another function would be to develop Parliament’s technical expertise so that MEPs could exercise more informed oversight of AI policies.

The report also sees a direct role for artificial intelligence in the legislative process. AI tools could be used to identify inconsistencies between acts, overlapping obligations, legal fragmentation, or situations in which national legislation adds requirements beyond the European standard.

AI could also contribute to analysing evidence, assessing existing legislation, and identifying areas where simplification is possible. However, the report establishes an explicit political limit: such tools must assist decision-making, not take it over.

Political decisions should not be made exclusively automatically, and responsibility for them must remain with the representatives and institutions entrusted with the power to decide.

Manolis Kefalogiannis summarised the approach by stating that the objective is for AI “to remain a tool in the service of people, democracies, and our values”.

AFCO’s adoption does not turn these proposals into Parliament’s definitive position. The report is due to reach the plenary during the 19–22 October session, where MEPs may confirm, amend, or reject the text.

https://2eu.brussels/ro/news/eurodeputatii-cer-control-uman-asupra-ai-si-reguli-internationale-pentru-protejarea-democratiei