Responsible AI for effective policy monitoring

Contrary to common prejudice, the EU law-making system is in many respects more transparent than those of many national governments — at least from a formal procedural perspective. Public records and proceedings, official documents, stakeholder contributions and expert opinions are publicly accessible through the EU’s extensive europa.eu domain, where the Union publishes vast amounts of information about the objectives, motives, research and influences behind its policies and laws.

Take for instance the parliamentary committees, where the core of policy work takes place: in the European Parliament, sessions are normally accessible to anyone via live stream, whilst the respective committees of national parliaments often meet behind closed doors even in some of the most transparent member states.

A sense of remoteness and complexity

Still, for policy professionals — let alone for ordinary citizens — EU institutions often appear more remote than their national governments. Indeed, geographically they usually are, but more importantly they receive less observant scrutiny in national media, which still tends to be country-specific and often reports on EU developments almost as if they were something closer to foreign affairs.

The sense of unfamiliarity is reinforced by the fact that the Union’s institutions and procedures differ from those of member states. From the names of political parties and factions to the language of legislative procedures and laws themselves, everything operates in a political idiom of its own. Politicians who are very influential within the Brussels bubble sometimes remain widely unknown outside of it.

An overwhelming volume of policy output

Above all, the policy-making process of a supranational community representing almost half a billion people across 27 countries is, by its very nature, extraordinarily complex. The EU has invested significant effort in improving its transparency and comprehensibility, but no matter how accessible each individual document or procedural detail may be, there are limits to how simple a political organism of continental dimensions with its more than 60 institutions, bodies and agencies can ever become.

To properly understand how EU policy comes about, it is nowhere near enough to grasp only the roles of the Commission, the Parliament and the Council. Decision-making involves numerous committees, dozens of working parties and thousands of expert groups. Some of the most critical procedures are not even fully formalised in the EU’s constitutional framework, such as the trilogues, whilst some of the most important laws are shaped through intricate proceedings like “comitology” or they take the form of peculiar instruments such as the “delegated acts”.

So, the real challenge for making sense of the EU’s inner workings does not mainly stem from a lack of formal transparency or access to documents. It lies in the sheer volume and complexity of procedures, policies and output. Over the years, EU competences have expanded into more areas of policy and law making. More member states and more official languages mean a greater diversity of interests at play. Cross-European and international interdependencies have deepened. And perhaps most importantly, the sheer number of legislative acts has risen.

During the legislature 2019–2024, the EU adopted more than 13,000 legal acts according to figures published by the Financial Times — a wave largely related to the "European Green Deal". In the current legislature many of these measures are already being reviewed, amended, or postponed, largely driven by the “Clean Industrial Deal”.

From monitoring to meaning: the evolving role of professionals

For any decision-maker affected by these laws and policies, staying aware and afloat — let alone ahead — is a formidable task. Leadership in organisations of all kinds therefore rely on in-house professionals or external consultants to cover their backs, alert them to policy changes and, ideally, contribute during the early phases of policymaking when influence is greatest. Yet even for seasoned professionals, staying on top of it all and shouldering the responsibility of not letting anything important slip through unnoticed has become harder than ever.

Over the decades, this challenge has given rise to a range of monitoring services: teams and platforms dedicated to following the policy agenda, tracking legislative procedures, transcribing and summarising live proceedings such as proposed laws and committee hearings in the European Parliament, and alerting clients to relevant developments.

But given that EU decision-making has grown to become so impactful and complex, it is no longer a topic for a narrow circle of specialists. It is a question for civil society at large: how to remain informed in order to comply with, respond to and participate in EU policymaking. Civil society here includes everyone from individual citizens, businesses, NGOs and interest groups to journalists, researchers, students, companies and industry alliances. Their experience with the EU — whether they feel able to understand and engage with it when they wish to — has a direct bearing on the commitment, confidence and support that societies extend to the European project as a whole.

AI as an enabler, not a substitute

From the viewpoint of policy professionals, the true significance of monitoring tools and services lies in their steady and reliable support for the most laborious and tedious routines of policy tracking: the never-ending accumulation of committee debates, midday press briefings, the publication of broad and narrow policy programmes, legislative initiatives, public consultations and calls for evidence, scrutiny periods for delegated acts, voting lists and results, thousands of amendment proposals, hours-long plenary debates and committee proceedings, trilogue outcomes, technical exchanges and an ever-growing stream of policy documents from the EU institutions and governments as well as individual politicians, think tanks, industry organisations and NGOs. In spite of all measures to increase transparency and accessibility, the actual challenge becomes how to distinguish the essential signals from the surrounding noise.

In this respect, artificial intelligence opens entirely new opportunities — not for substituting human intelligence and discretion, but for augmenting people’s capacity to absorb and make sense of vast amounts of information that far exceed human capacity for ingestion. It enables the detection of critical details buried like a pinhead in an entire hayfield, and the timely delivery of those insights to the people to whom they matter. It also helps to detect opportunities for participation — such as public consultations, calls for evidence and stakeholder feedback rounds — in order to engage at precisely the moments when input has the greatest value for policy outcomes. It helps focus resources on what truly matters, empowering human judgment, closing blind spots and supporting better-informed decisions — not only in the interest of individual users, but of society as a whole.

Responsible design and regulation

To make the most of that potential, such systems must be designed with care. They need to be accurate, reliable, resilient against manipulation and disinformation, timely and context-aware. Just as importantly, they must be accessible and broadly affordable; otherwise, they risk deepening rather than reducing informational divides in society. AI makes all this possible — but only through careful, diligent and advanced technological approaches as well as responsible design.

A responsible approach to artificial intelligence in policy monitoring begins with the principle of supporting human sovereignty, intelligence and discretion rather than replacing it. Artificial intelligence can help individuals and societies focus on what truly matters by cutting through the fog of information overflow. Personalisation is essential: users should be able to receive information aligned with their specific priorities and areas of responsibility.

Maintaining high standards of quality control is equally important. Responsible systems follow rigorous safeguards against misinformation, manipulation and offensive content, while continuously improving their procedures and encouraging users to report inaccuracies or issues. Neutrality, impartiality and transparency must guide all stages of development and use, including explicit disclosure when and how AI is employed. And of course, compliance with the EU’s Artificial Intelligence Act must be ensured.

A glimpse into practice: The EU Commission hearings

A recent example illustrates how such a system can support not only individual users but also analysts, journalists and researchers who aim to provide timely and accurate information to the wider public. During the current European Commission’s confirmation hearings — twenty-six sessions held over several days, comprising more than eighty hours of multilingual material, often aired simultaneously — our system was able to process in near real time what would otherwise have required days of manual effort.

Within minutes of each hearing’s conclusion, transcripts and summaries were available and fully searchable, both through keywords and natural language queries. Users could, for example, ask “What is the candidate’s position on wetlands and forests?” and instantly receive an answer – if necessary – compiled from relevant statements made at different points in the session. Extracting such insights is now a matter of seconds, enabling users to make sense of information far more efficiently and to dedicate their time to higher-value human intelligence output: interpreting meaning, drawing conclusions, identifying implications in the context of human societies — and ultimately holding policymakers to account for what they have said and committed to before any democratic decisions on appointments or legislation are made.

The future of policy monitoring

The future of policy monitoring must not be defined by whether machines can replace human judgment, but by how effectively they can strengthen it. Artificial intelligence can make policymaking more transparent, responsive and intelligible, but only if applied responsibly and with broad accessibility in mind. The real progress lies in freeing human intelligence from mechanical routine, allowing policy professionals, analysts, journalists, researchers and citizens alike to focus on meaning rather than drowning in mountains of information. In that sense, the next generation of policy monitoring is not simply about keeping up with the pace of the EU — it can be a contribution that Europe’s democratic processes remain understandable, accountable and open to all who wish to follow and participate in them.

Stay in the loop with Thembi

We are building a smarter way to track and understand EU policy, combining AI-powered monitoring with intuitive insights that deliver fast, comprehensive and actionable intelligence.

Thembi helps you stop chasing data and start driving policy work.

Sign up to be the first to know when we launch.