Recommendation 4: The mitigation of risks to pluralism and democratic processes posed by AI systems should be a priority area in the implementation of

Although the AI Act took effect in August, key provisions—especially those concerning GenAI systems whether text, image or multimodal—won’t apply for another one to two years. In the meantime, the GPAI CoP will be critical for guiding GenAI governance. The Commission must ensure that the AI Act and GPAI CoP implementation addresses more than just extreme risks, such as the production of ABC weapons or cyber-attacks. Social and informational risks, such as hallucinations, hate speech, bias, and discrimination, must also be tackled, as well as the broader impact of GPAI systems on the information ecosystem before, during and after elections. Incorporating AI risk management into the Commission's AI innovation strategy is essential to create a safer, sustainable and more trustworthy AI industry.

• Ensuring AI Accountability: Auditing Methods to Mitigate the risks of Large Language Models. (October 2024)

• AI Act Comes into Force: What It Means for Elections and DRI’s Next Steps. (August 2024)

• Are Chatbots Misinforming Us About the European Elections? Yes. (April 2024)

The Union has taken significant action through the DSA, but these now need to be taken forward with effective enforcement and continuous monitoring of impact. This should cover more marginal platforms, such as Rumble, Odyssey, Truth Social and Telegram, which are often more radicalised but are not considered to be VLOPs. The implementation and monitoring of the European Media Freedom Act should also ensure that it tackles the lack of transparency of media ownership, and the lack of effective independence in the media regulators of some member states. See also:

• Monitoring media pluralism in the digital era: Application of the media pluralism monitor in the European member states and in candidate countries in 2023, European University Institute.