Right to challenge AI decisions

What it is

The right to challenge AI decisions is the right to contest outputs, recommendations or decisions produced by artificial intelligence systems when you believe they are:

• wrong (a technical error, incorrect data);
• unfair (they ignore the context or particular circumstances);
• discriminatory (algorithmic bias on gender, age, ethnic origin and more);
• harmful to your rights.

What the AI Act provides

A complaint to the authority (Art. 86)

Natural persons who believe the AI Act has been infringed can lodge a complaint with the competent authority:

• the national AI supervisory authority;
• in Italy: AGCOM (coordination), the Data Protection Authority and the other sector authorities.

The authority must:

• examine the complaint;
• investigate the infringement;
• inform you of the outcome;
• take appropriate measures (penalties, compliance orders).

Going to court (Art. 85)

If the complaint is rejected or ignored:

• you can sue the provider of the AI system;
• the court can annul the decision, award damages, or order the system to be withdrawn.

How it differs from the DSA internal complaint (Art. 20)

DSA complaint (online platforms):

• challenges content moderation;
• an internal complaint to the platform;
• covers moderation decisions.

AI Act complaint (high-risk AI systems):

• challenges algorithmic decisions with a significant impact;
• a complaint to a public authority;
• covers any decision by a high-risk AI system.

The two remedies can overlap: if a platform uses AI to moderate content, you can use both.

AI decisions you can challenge

Employment

The AI rejected your application. Grounds for challenge:

• it discarded your CV for missing keywords, even though you had the skills;
• it scored your video interview negatively because of bias (accent, gender, age);
• it applied rigid criteria unsuited to the role.

The AI rated your performance negatively. Grounds:

• it measures productivity with quantitative metrics and ignores quality;
• it does not consider external factors (illness, varying workloads);
• a low rating exposes you to the risk of dismissal.

The AI denied you a promotion. Grounds:

• it classified you as “not ready” without considering your achievements;
• it is biased against under-represented groups;
• its criteria are not transparent.

Credit

The AI rejected a mortgage or loan. Grounds for challenge:

• the credit score is wrong (outdated or incomplete data);
• it does not consider the recent improvement in your finances;
• it applies generic penalties that do not fit your case.

Education

The AI gave you a low grade. Grounds for challenge:

• it marked the exam with rigid criteria, without assessing your reasoning;
• it penalised form (grammar) and ignored content;
• it does not recognise equivalent answers phrased differently.

The AI denied you university admission. Grounds:

• it only assessed grades, ignoring extracurricular activities and motivation;
• it is biased towards prestigious schools and penalises those from lesser-known ones;
• its criteria are not transparent or not appropriate.

Public services

The AI denied you a benefit. Grounds for challenge:

• it only assesses formal economic parameters, ignoring real hardship;
• it does not consider necessary expenses (medical care, caring for relatives);
• its decision contradicts the social worker’s assessment.

The AI gave you low priority for public housing. Grounds:

• it uses an automated score that unfairly penalises you;
• it does not consider urgency (eviction, housing emergency);
• other people in similar or worse situations have higher scores.

Justice

The AI assessed you as at high risk of reoffending. Grounds for challenge:

• it uses discriminatory factors (neighbourhood, social background);
• it does not consider your rehabilitation and change of life;
• a high score leads to harsher precautionary measures or the denial of parole.

Reinforced rights:

• challenging it with legal assistance;
• the judge must give reasons if they follow the AI’s recommendation;
• the right to a fair trial prevails over the AI’s output.

Healthcare

The AI suggested an inappropriate diagnosis or treatment. Grounds:

• it analysed medical images but misinterpreted them;
• it suggests a standard protocol unsuited to your comorbidities;
• it does not consider your full medical history.

Your right to challenge:

• a human second medical opinion;
• additional tests;
• a new diagnostic assessment.