Back to Administration of justice

The principles identified in the Framework Convention Framework Convention (Articles 4 to 13)  and the European Ethical Charter on the Use of Artificial Intelligence correspond to significant, real concerns vis-à-vis the use of AI in administration of justice and its possible negative impacts on of human rights as protected in the ECHR, as well as in Convention 108(+). Principles of the Ethical Charter include respect for fundamental rights, non-discrimination, quality and security, transparency, impartiality and fairness; and the principle of “under user control”. The principle of “under user control” precludes a prescriptive approach and ensuring that users are informed actors and in control of their choices.

The human right primarily impacted in this sector is the right to a fair trial, guaranteed by Article 6 ECHR. Also other international human rights instruments (articles 10 and 11 of the Universal Declaration of Human Rights, article 14 of the International Covenant on Civil and Political Rights, article 47 of the Charter of Fundamental Rights of the European Union, article 8 of the American Convention on Human Rights-Pact of San José, article 7 of the African Charter of Human and Peoples’ Rights) and in the constitutional legal order of democratic countries.The right to liberty and security, guaranteed by Article 5 ECHR, is also engaged insofar as it involves special rules on judicial protection against arbitrary deprivation of liberty.

The right to a fair trial 

Article 6 ECHR applies to proceedings involving the determination of civil rights and obligations or of any criminal charge. The key principle governing Article 6 is fairness.[1] As highlighted by the Court, what constitutes a fair trial cannot be the subject of a single unvarying rule but must depend on the circumstances of each case and in light of the overall fairness of the proceedings.[2] Certain subsidiary principles of fairness are also particularly relevant in the AI context:

Independence and impartiality

Article 6 guarantees a fair and public hearing by an independent and impartial tribunal established by law.[3] The tribunal should be independent both from other branches of government, such as the executive and legislature, and from the parties involved in a case.[4] The tribunal must also be impartial, namely subjectively free of personal prejudice or bias and must offer sufficient guarantees to exclude any legitimate doubt in this respect.[5]

Bias in AI systems may not be easily discernible by the judge due to the generalised perception of algorithmic/mathematic “neutrality” and judges’ own technology bias. This could lead to discriminatory outcomes. Extensive reliance on AI could lead to a “standardisation” of judicial decisions, with judges feeling compelled to follow AI recommendations due to the perceived “superiority”, particularly in systems where their terms of office are not permanent but subject to popular vote,[6] or in which their personal liability (disciplinary, civil or even criminal) is likely to be incurred.[7]

Back to top
 

Presumption of innocence

The principle of presumption of innocence in criminal proceedings requires, among other things, that: (i) judges (and jurors where applicable) must approach their duties without any preconceived notion of the accused's guilt; (ii) the burden of proof is on the prosecution, and (iii) any doubt should benefit the accused.[8]

As a result of algorithmic bias, the potential inclusion in AI systems of variables such as criminal history and family background means that the outcome of an individual’s case may be affected by the past behaviour of a certain group without appropriate attention to the accused individual’s specific background, motivations and, eventually, guilt. This could result in interfering with a person’s right to be presumed innocent until proven guilty by a court of law. While the use of predictive tools by judges in criminal trials is very rare in Europe,[9] in other jurisdictions there are real-life examples of the negative effects.[10]

Back to top
 

Equality of arms and adversarial proceedings

Equality of arms is an inherent feature of a fair trial. It requires that each party be given a reasonable opportunity to present a case on conditions that do not place him or her at a disadvantage vis-à-vis the opponent and applies to criminal and civil proceedings.[11] In a criminal context, the right to adversarial proceedings further means that the accused have the opportunity to familiarise themselves with and to comment on all evidence adduced or observations filed with a view to influencing the court’s decision, its existence, contents and authenticity in an appropriate form and within an appropriate time.[12] Failure to disclose to the defence material evidence which could enable the accused to exonerate themselves or have their sentence reduced would constitute a refusal of facilities necessary for the preparation of the defence, and therefore a violation of Article 6.[13] The right to adversarial proceedings may not be disregarded to save time and expedite the proceedings.[14]

In the context of the use of AI systems, equality of arms suggests that in every case, the party should be aware of the role of the system in the proceedings, the criteria for its operation and the possible impact on the outcome of the case. Concerns may arise if a party is denied sufficient access for scrutiny of AI-analysed data used as evidence.[15] The right to adversarial proceedings is likely to require the ability to challenge an AI system's scientific validity, biases, and potential errors. However, intellectual property rights and trade secret laws may restrict access to privately-owned proprietary AI systems used by law enforcement authorities. Even without these obstacles, the complexity of the models used (“the black box problem”) may present a major challenge for the defendant. Furthermore, while AI systems may expedite proceedings by saving time, the right to adversarial proceedings cannot be disregarded for this purpose.

In civil proceedings, equality of arms could be undermined by a possible imbalance between the parties to the dispute in their understanding and ability to use AI tools, with respect to their available means, including financial means, or even their digital literacy level. In that context, Recommendation CM/Rec(2016)3 of the Committee of Ministers to member States on human rights and business highlights that when alleged victims of business-related human rights abuses bring civil claims related to such abuses against business enterprises, member States should ensure that their legal systems sufficiently guarantee an equality of arms within the meaning of Article 6 of the ECHR. In particular, they should provide in their legal systems for legal aid schemes regarding claims concerning such abuses. Such legal aid should be obtainable in a manner that is practical and effective.[16] In this context, the AI tools could also be developed to provide legal information and assistance and may facilitate seeking human rights protection by users, but again, appropriate safeguards should be put in place when designing such schemes.

Back to top
 

Access to court

The right of access to a court is an inherent aspect of the safeguards enshrined in Article 6 and is no more absolute in criminal than in civil matters. Everyone has the right to have any claim relating to his “civil rights and obligations” brought before a court or tribunal.[17] An individual must have a clear, practical opportunity to challenge an act that is an interference with his rights.[18] The practical and effective nature of this right may be impaired by, for instance, excessive formalistic interpretation of procedural rules.

Within that context, resorting to AI systems, should not hinder the right of access to a court within the meaning of Article 6[19] ; human oversight over decision-making should be ensured.[20] Access to court should also not be hindered by technical hurdles related to a specific AI system. In that respect, by not considering the practical obstacles linked to the required use of an e-filing system and by not allowing for alternative (paper) submission, a domestic court may be taking an excessively formalistic approach amounting to a violation of Article 6§1.[21]

Right to liberty and security

The key purpose of Article 5 is to prevent unlawful, arbitrary or unjustified deprivations of liberty.[1] In order to meet the requirement of lawfulness, deprivation of liberty must be in accordance with a procedure prescribed by law and subject to review of its lawfulness by a court. While flaws in a detention order do not automatically render the detention unlawful,[2] the reasoning of the decision ordering a person’s detention is a relevant factor in determining whether the detention must be deemed arbitrary under Article 5 § 1.[3] 

Deprivation of liberty is also unlawful under Article 5 if it results from a conviction following proceedings which were “manifestly contrary to the provisions of Article 6 or the principles embodied therein”, to the point of constituting a “flagrant denial of justice”.[4] A trial that is summary in nature and does not allow for a thorough and objective assessment of the case of a person who is then sentenced to imprisonment could thus amount to a violation of not only Article 6, but also Article 5 ECHR[5].

Issues arising under Article 6 (see above) will also apply to the judicial guarantees against arbitrary detention set out in Article 5 §§ 3 and 4, which respectively provide that criminal detainees shall be brought promptly before a judge and that everyone may challenge the lawfulness of their detention before a court. Lack of transparency or accountability in potential AI-systems could undermine the fairness of decisions on deprivation of liberty. They risk perpetuating biases, leading potentially to unjust pre-trial detention, disproportionate sentencing, or unfair parole denials. Additionally, their opacity challenges individuals’ ability to contest decisions effectively, raising concerns about fairness and accountability.

Back to Administration of justice