Responsible AI Platform

High-Risk AI in Justice and Democratic Processes

··5 min read

Domain 8 of Annex III AI Act covers two fundamentally different applications: AI in the administration of justice and AI influencing election outcomes. Both touch pillars of democratic rule of law. The Commission guidelines of 19 May 2026 draw sharp boundaries here.

For the general framework, see the main article on the Article 6(3) filter. For all domains, see the hub overview.

The Two Use Cases of Domain 8

  • Point 8(a) AI systems supporting judicial authorities or alternative dispute resolution
  • Point 8(b) AI systems intended to influence the outcome of elections or referendums, or voting behaviour of natural persons

Point 8(a): AI in the Judiciary

For the judiciary and alternative dispute resolution (arbitration, mediation), any AI system substantively contributing to investigation or interpretation of facts and law is high-risk.

High-risk:

  • AI analysing facts in a case file and generating recommendations
  • AI conducting legal analysis for judicial decisions
  • AI searching and interpreting case law in concrete matters
  • AI conducting analyses in mediation or arbitration influencing the outcome

Outside scope:

  • Pure administrative assistance: calendar management, file management, forms
  • Spelling and grammar checking of judgments by judges
  • Transcription services for hearings
  • Anonymisation tools for publication of judgments

The guidelines make clear the boundary lies at "substantive contribution to investigation or interpretation of facts and law". Pure supportive technology is excluded, but as soon as the system substantively contributes to legal reasoning, you are in scope.

Point 8(b): Electoral Influence

This is the most politically charged use case of Annex III. The guidelines explain sharply what does and does not fall within it.

High-risk:

  • AI specifically intended to influence election outcomes through microtargeting
  • AI generating synthetic media (deepfakes) for election purposes
  • AI influencing individual voters through personalised persuasion messages
  • AI developing strategies to demobilise specific voter groups

Outside scope:

  • Legitimate campaign organisation (call lists, volunteer management)
  • Polling and sentiment analysis at aggregate level
  • General political communication and advertising (covered by other regimes such as the Digital Services Act and the Political Advertising Regulation)

Important distinction: General political ads optimisation under the Political Advertising Regulation does not automatically fall under Point 8(b). But as soon as the system is specifically intended to influence the outcome of an election, you are in scope.

Sector-Specific Pitfalls

Pitfall 1: Legal Tech and Judicial Users

Many legal tech tools are sold to both lawyers and judges. For lawyers it falls outside Annex III. For judicial users it falls under Point 8(a). The same software can therefore behave differently depending on who deploys it.

Pitfall 2: Arbitration and Mediation Often Forgotten

Point 8(a) explicitly names alternative dispute resolution. For arbitration institutes and mediation platforms deploying AI, the same rules apply as for the judiciary.

Pitfall 3: Synthetic Media on Top of Article 50

For synthetic media, the transparency obligations of Article 50 AI Act also apply, plus possibly the Digital Services Act for very large online platforms. A deepfake in a political campaign touches three regimes simultaneously.

What to Do

1

For the judiciary: separate administrative from substantive support

Calendar, archival, anonymisation usually fall outside. Facts and law analysis usually fall within.

2

For legal tech vendors: know your deployer

Sales to judges or arbitration institutes places your system in high-risk regime. Build conformity assessment and documentation for that.

3

For political campaigns and media: check Article 5 and Article 50

Manipulative techniques can already fall under Article 5 prohibition. Synthetic media disclosure under Article 50 is mandatory in any case.

Practical tool

Test your AI against the Article 6(3) filter

Free interactive self-assessment, updated for the Commission guidelines of 19 May 2026. 9 steps, personal report with reasoning, vendor questions and next steps.

Start the Annex III Classifier 2026Free · 5-8 minutes · Article 6(3) filter built in

Frequently Asked Questions

⚖️ Referenced Legislation

High-Risk AI in Justice and Democratic Processes | Practical Guide | EU AI Act | Responsible AI Platform