Authored By: Arwa Yasir
Middlesex University Dubai
Introduction
The increasing use of artificial intelligence (AI) in significant legal decision- making raises an important challenge for criminal justice systems: should AI systems assist judges in determining sentencing decisions? The rapid development of Al technology has generated opportunities for improved consistency, efficient, and evidence- based decisions in sentencing. However, sentencing requires careful evaluations of individual circumstances, proportionality, rehabilitations, and responsibility, which pose significant challenges for automated decision-making.
In England and Wales, sentencing remains a judicial function governed by the Sentencing Act 2020 and guided by Sentencing Council guidelines.[1] These legal frameworks state that courts use discretion while taking the offender’s circumstances and the significance of the offence. Although AI systems can help with judicial decisions by identifying patterns and offering risk assessments, there are many complications that come along with these advancements.
Judicial sentencing decisions should be supported by artificial intelligence, not replaced by it. Without proper safeguards, it could jeopardize justice, transparency, and equality under the law. This article focuses on the UK sentencing procedure, the potential benefits of AI-assisted sentencing, and the legal and ethical dangers associated with algorithmic decisions.
Legal Framework Governing Sentencing in England and Wales
In England and Wales, sentencing decisions are founded on the principle that punishment should be proportionate to the severity of the act and the offender’s level of guilt. The Sentencing Act 2020 provides the legal basis for sentencing by mandating courts to comply with appropriate sentencing guidelines unless doing so would be contrary to the interests of justice.[2] In order to maintain judicial discretion and encourage consistency, the Sentencing Council issued sentencing guidelines.[3]
Therefore, the present judicial framework relies on human assessment. Judges are required to take into account a variety of criteria, such as potential for rehabilitation, personal circumstances, prior convictions, offender liability, and harm caused. These standards require contextual judgment rather than any simple categorizations.
The application of AI in sentencing sets individual justice and consistency at conflict. AI conflict could analyse vast amounts of sentencing data and uncover patterns that could assist judges. However, moral and legal considerations are involved in sentencing judgments, which are difficult to reduce to scientific predications.
Based on Ryberg, the “input problem” is one of the main challenges with AI sentencing tools.[4] If an algorithm generates penalty recommendations based on previous judicial decisions, the system may duplicate the assumptions and disparities identified in those rulings. The standard of an AI assessment is determined by the accuracy and fairness of the information used to train the system.
This raises a significant legal issue. Applying legal principles to specific conditions is just as crucial in sentencing as predicating outcomes. A system relying mainly on statistical patterns undermining the requirement that sentencing decisions remain fair and proportionate.
Opportunities of Artificial Intelligence in Criminal Sentencing
Artificial intelligence has the potential to improve various aspects of criminal sentencing. Increased consistency is one potential benefit. Sentencing decisions may differ depending on how judges view similar cases. AI technologies could give judges comparative sentencing data to determine whether suggested penalties are consistent with earlier rulings.
Ryberg argues that AI could improve sentencing performance by supporting courts in reviewing massive datasets and discovering important elements influencing sentencing verdicts.[5] AI could function as an analytical tool to support judicial decision-making rather than replacing the role of judges entirely.
However, AI systems must adhere to legal restrictions. Under Article 6 of the European Convention on Human Rights, individuals are entitled to a fair hearing before an independent and impartial tribunal. The right to a fair trial before an impartial, independent tribunal is guaranteed by Article 6 of the European Convention on Human Rights.[6] A sentencing system that relies only on autonomous decision-making raises concerns over whether the accused is given significant judicial consideration.
Recent legal developments reveal that judges are becoming more conscious of the dependability of AI. Although Chandra v Royal Mail Group did not concern criminal sentencing, the case emphasized the judicial caution when using AI-generated material and its importance of verifying technological outputs before completely relying on them in legal proceedings.[7] It highlights a broader judicial concern that AI-generated information cannot always be regarded as trustworthy and accurate.
Similarly, M v. F (Fact Finding Hearing) addressed challenges with AI-generated evidence and the requirement of carefully assessing technological material before they are relied upon and presented in court.[8] These rulings imply that courts are acknowledging AI as a helpful but potentially flawed tool that requires intensive human assessment.
Risks of Algorithmic Bias and Reduced Judicial Accountability
The potential for discrimination through algorithmic bias is the major concern with AI sentencing. Biased previous judgments may have an impact on future guidelines since AI systems gain insight from existing datasets. Arowosegbe argues in ‘Data Bias, Intelligent Systems and Criminal Justice Outcomes’ that AI systems used in criminal justice are trained on data that reflects present patterns of policing, prosecution, and punishment, and they may reinforce existing inequalities already established.[9]
Due to the direct effect of judicial decisions on liberty, this raises serious concerns. If the algorithm suggests harsher penalties for certain groups based on unreliable precedents, the system may produce inconsistent outcomes while appearing objective.
Another key issue is transparency. Many AI systems use complex algorithms that are hard for humans to comprehend. An offender should be able to comprehend the reasoning behind a punishment that was determined by an algorithm. Procedural fairness standards can conflict with a lack of transparency.
Similar challenges are presented by Shi’s analysis of AI’s application in Chinese criminal justice.[10] He claims that when technology influences judicial decisions, accountability becomes ambiguous, which makes algorithmic sentencing methods controversial. Criminal justice systems require clear accountability for decisions affecting individual liberty.
Concerns regarding AI dependability are also reflected in the US judicial decisions. The United States District Court examined problems with AI-generated content and the risks of unreliable technological outputs in Flycatcher Corp Ltd and Flycatcher Toys Inc v. Affable Avenue LLC.[11] Although this case is unrelated to criminal sentencing, it reinforces the broader principal that judges should exercise caution when exercising AI-generated material.
Therefore, Artificial Intelligence should not be used to make final decisions in criminal sentences. Judges should make the final judgment due to the legal interpretation, ethical judgment, and consideration of personal circumstances, which computers cannot entirely comprehend.
Conclusion
In England and Wales, artificial intelligence offers both advantages and disadvantages for criminal punishment. By helping courts with complex data analysis, AI systems could increase uniformity, efficiency, and access to sentencing information. However, sentencing serves as a human legal judgment that considers each case’s particular circumstances along with proportionality and fairness.
Therefore, AI should be used in supporting judicial decisions rather than replacing the discretion exercised by the judges. The concerns of algorithmic prejudice, a lack of transparency, and limited accountability establish that complete digitized sentencing will violate key criminal justice principles.
The establishment of a regulated framework for AI-assisted sentencing should be the main objective of future legal development. Transparency, human supervision, and continuous evaluation of algorithmic accuracy should all be part of this model. AI can remain as a tool in the justice system, while judges must continue determining sentencing judgments.
Bibliography
Cases:
Chandra v Royal Mail Group (Case No 3311062/2023, Employment Tribunal, 25 July 2025).
Flycatcher Corp Ltd and Flycatcher Toys Inc v Affable Avenue LLC (24 Civ 9429 (KPF), SDNY, 5 February 2026).
M v F (Fact Finding Hearing) [2026] EWFC 22.
Legislation:
European Convention on Human Rights 1950.
Sentencing Act 2020.
Secondary Sources:
Jesper Ryberg, ‘Criminal Sentencing and Artificial Intelligence: What is the Input Problem?’ (2025) 19(2) Criminal Law and Philosophy
Jesper Ryberg, ‘Criminal Justice and Artificial Intelligence: How Should We Assess the Performance of Sentencing Algorithms?’ (2024) 37(1) Philosophy & Technology
Jacob O Arowosegbe, ‘Data Bias, Intelligent Systems and Criminal Justice Outcomes’ (2023) 31(1) International Journal of Law and Information Technology
Jiahui Shi, ‘Artificial Intelligence, Algorithms and Sentencing in Chinese Criminal Justice: Problems and Solutions’ (2022) 33(2) Criminal Law Forum 121.
Other Sources:
Sentencing Council, Sentencing Guidelines https://www.sentencingcouncil.org.uk accessed 09 July 2026.
[1] Sentencing Act 2020, ss 59–60; Sentencing Council, Sentencing Guidelines.
[2] Sentencing Act 2020, s 59.
[3] Sentencing Council, Sentencing Guidelines.
[4] Jesper Ryberg, ‘Criminal Sentencing and Artificial Intelligence: What is the Input Problem?’ (2025) 19(2) Criminal Law and Philosophy 203.
[5] Jesper Ryberg, ‘Criminal Justice and Artificial Intelligence: How Should We Assess the Performance of Sentencing Algorithms?’ (2024) 37(1) Philosophy & Technology 9.
[6] European Convention on Human Rights 1950, art 6.
[7] Chandra v Royal Mail Group (Case No 3311062/2023, Employment Tribunal, 25 July 2025).
[8] M v F (Fact Finding Hearing) [2026] EWFC 22.
[9] Jacob O Arowosegbe, ‘Data Bias, Intelligent Systems and Criminal Justice Outcomes’ (2023) 31(1) International Journal of Law and Information Technology 22.
[10] Jiahui Shi, ‘Artificial Intelligence, Algorithms and Sentencing in Chinese Criminal Justice: Problems and Solutions’ (2022) 33(2) Criminal Law Forum 121.
[11] Flycatcher Corp Ltd and Flycatcher Toys Inc v Affable Avenue LLC (24 Civ 9429 (KPF), SDNY, 5 February 2026).
