From Legal Jargon to Citizen Voice: AI-Enhanced Tools for Informed Consent and Participatory Digital Regulation

A new Policy Brief by Work Package 3 presents initial findings from the ongoing research into a central challenge of digital regulation: informed consent often fails to work in practice.

Terms of service for digital services often number several thousand words, leading to the vast majority of users do not read them in their entirety. This raises the question of how legal information can be communicated in a way that allows people to make truly informed decisions.

AI for More Understandable Legal Texts

As part of WP3, an AI-enhanced tool was developed that translates complex legal and regulatory language into plain language at B1 level. The tool aims to make legal information more accessible while preserving its original meaning. 

In order to achieve this, the policy brief identifies three key requirements:

• Faithfulness: Every summary must preserve the legal meaning of the underlying clause with no material terms omitted or softened.

• Readability: The output must be understandable for a general audience without specialist background.

• Source Traceability: Every simplified point must link directly back to the source clause, so both users and regulators can verify it.

Where is the line between summarization and legal advice?

A key challenge lies not only in the technical implementation but also in the legal classification of such systems.

Using the German Legal Services Act (RDG) as an example, this policy brief examines the distinction between permissible summarization or translation and legal services that may require authorization. While the clear presentation of legal texts must generally be distinguished from an individual legal assessment, the line can become relevant when a tool evaluates specific legal consequences for an individual case or provides recommendations for action.

This creates a dilemma for developers: The scope of functionality must be designed to comply with legal requirements—yet it is precisely these more advanced functions that are potentially particularly helpful to users.

Between Comprehensibility and Decision-Making Autonomy

From an ethical perspective as well, the simplification of legal language raises questions. Every summary and simplification involves decisions about which information to highlight, simplify, or omit.

This policy brief examines this conflict of objectives in light of the guidelines for trustworthy AI: More understandable information can improve access to legal content, but at the same time, the presentation must not lead to users being influenced in their decisions by a selective portrayal.

Policy Recommendations:

Based on the findings to date, the policy brief identifies four areas for action:

• Strengthen feedback: Structured channels between citizens and regulatory authorities can help incorporate practical experiences with digital regulation more systematically.

• Ensure legal certainty: Greater clarity is needed regarding which functions of AI-supported legal assistance are permissible and where the line is drawn with respect to regulated legal services.

• Review compliance requirements: The proportionality of existing regulatory requirements should be examined, particularly with regard to innovation and smaller providers in the legal tech sector.

• Develop nuanced guidance for LLM applications: EU regulation should distinguish between different risk levels and uses of LLMs, particularly between tools for simplification or summarization and those providing advice. Ethical and legal assessments should also be better aligned to ensure consistency with frameworks such as the EU AI Act.

The results should be understood as preliminary findings from ongoing research. As the AI4POL project progresses, the analyses will be deepened and further developed.