The use of artificial intelligence in the Mexican justice system
Artificial intelligence (AI) consists of the ability of machines or computers to emulate human thinking and decision making. A key area within AI is machine learning. With AI platforms and algorithms and machine learning, you can measure the happiness levels of each individual, perform certain actions to improve the experience with the user and even build a judicial or penitentiary system with very specific analysis and classification tasks.
In Mexico, there is a need for more efficient prison security, especially in terms of managing infrastructure, physical and logical security, and also people.
Another issue that sets a challenge in Mexico’s federal prison system has to do with the slowdown and lack of effectiveness in the processes of prisoner classification and risk assessment. In addition, in the judicial sector, the country lacks an automatic analysis platform that supports jurisprudence.
User and Entity Behaviour Analytics (UEBA) is the monitoring, collection and evaluation of data and activities of users who interact with a system which may be related to information, transactions or processes.
UEBA technologies use AI and machine language to analyse historical data records (Big Data), which can contain numbers, text, voice, audio and video; to identify patterns and feed systems that support decision-making in fields such as the classification of individuals, social reintegration, physical security, logical security and cybersecurity. These systems can take measures or actions based on their findings and can be automatically adjusted to the systems to make “intelligent automated decisions”.
User behaviour analysis tools have more advanced profile and exception monitoring functions than computer systems and are used to determine a baseline of normal activities specific to the organisation and its individual users, and to identify deviations from the norm. UEBA uses big data and machine language algorithms to evaluate these deviations in near real time. This type of analysis allows an organisation to make classifications, make decisions, detect non-visible patterns and discover risk situations or other potential security threats.
UEBA collects various types of data, such as the roles and titles of users, including access, accounts and permissions; user activity and geographic location; and security alerts. These data can be collected from past and current activities, and the analysis considers factors such as the resources used, the duration of the sessions, the connectivity and activity of the peer group in order to compare anomalous behaviours. It is also automatically updated when changes are made to the data, such as added permissions.
UEBA systems do not report all anomalies as risky. Instead, they evaluate the potential impact of the behaviour. If it involves less sensitive resources, it receives a low impact score; if it is something more sensitive, such as personal identification information, you will receive a higher impact score. In this way, security teams can prioritise which tracking to follow, while the UEBA system automatically restricts or increases the authentication difficulty for the user who shows an anomalous behaviour.
Automatic learning algorithms allow UEBA systems to reduce false positives and provide clearer and more accurate actionable risk intelligence to cybersecurity teams.
In Mexico, tests are being carried out with the UEBA systems in three areas within the justice system:
More efficient prison security: through the public-private partnership models developed to build the prison infrastructure that began in Mexico ten years ago, private operators have begun to test AI platforms and machine language to manage the infrastructure of general services, security -both physical and logical- and the people who work daily (there are more than one hundred thousand devices and thirty thousand people).
Intelligent classification: Mexico’s federal prison system has tested to be able to rely on AI tools for carrying out classification processes more quickly and efficiently. Tests have been developed to predict whether the defendants show a risk of escape through their criminal and court records. Another field where tests have been successfully carried out is that of prevention, where a council has put these tools to the test in order to support the administration of the risk of a person committing another crime (recidivism).
Automated litigation analysis platform: through this platform, the judiciary is testing an AI system that, when hearing a plea or query automatically, reviews a database of millions of records that include legal citations, suggests articles to review and even calculates a confidence level, to help lawyers prepare cases. In addition, the more queries it receives, the more it learns, and its effectiveness increases. With these algorithms, you can take into account the opinion of the judge, the parties involved in the trial and the courts. Once it processes the information, it responds based on the current laws and translates the terminology. The platform tracks, in real time, the results of verdicts and judgements that have established jurisprudence, so that it can warn of any risk that represents a threat to its clients and correct it. The system interprets, executes value judgements according to a meticulous study on jurisprudence and stores them in a repository of structured and unstructured data, following the trends of data architectures and analysis for big data.
In terms of prison security, the AI platform is intended not only to make the verification teams more efficient but also the logistics of visits, verifications and attention to events that are really necessary. This prevents the development of a purely vicious circle that can even lead to situations of corruption or jailbreaks. It has resulted in a benefit in both an administrative sense (cost-benefit) and an operational sense (better planning of attention to defects versus the daily internal operation of prison facilities). Today, teams of people visiting these facilities are available to check or verify the current state of the infrastructure.
The intelligent classification tools resulted in being more accurate than the judges in predicting what the defendants would do after being released, while thanks to the prevention tools, the system reduced the crime rate in a controlled way.
Jorge Medina is a Global Security Expert at TUTUM Tech, Mexico. He has over 20 years’ experience in multinational tech companies where he was responsible for leading solutions and best practices for Governments across the Americas. He was also the CIO for the Mexico federal government in public safety and national security areas for many years. Mr Medina holds a degree in Electronic Engineering and Public Safety, a Diploma in Certified Information Systems Security Professional) and an MBA.