Responsible Use of AI in Policing

Key initiatives to ensure the responsible and ethical use of AI in policing.

"AI is undeniably a game changer for criminals and law enforcement alike. However, it is imperative that we make the shift to the new technological era in a trustworthy, lawful and responsible manner, providing a clear, pragmatic, and most of all useful way."

INTERPOL Secretary General Jurgen Stock

Analyst Perspective

Striking the balance between leveraging technology for public safety and protecting individual rights and freedoms.

The responsible use of artificial intelligence (AI) in policing and public safety is a multifaceted issue that encompasses several critical areas: data privacy, safety and security, explainability and transparency, fairness and bias detection, validity and reliability, and accountability. Each of these areas presents its own set of challenges and necessitates specific initiatives to ensure that AI technologies are used ethically, effectively, and in a manner that respects individual rights and promotes public trust.

The responsible use of AI in policing requires a comprehensive approach that addresses these critical areas through continuous improvement, stakeholder engagement, and adherence to ethical, legal, and societal standards. By tackling the challenges and implementing suggested initiatives presented in this research, law enforcement agencies can leverage AI technologies to enhance public safety while respecting privacy, ensuring security, and promoting fairness and transparency.

Neal Rosenblatt
Principal Research Director
Public Health Industry
Info-Tech Research Group

Executive Summary

Your Challenge

• AI integration in policing faces multifaceted challenges impacting its effectiveness and ethical implementation.
• Ensuring AI systems avoid discriminatory outcomes and address inherent biases is a pressing challenge.
• Balancing the needs for effective law enforcement with individuals' right to privacy remains a complex issue.
• Determining responsibility and accountability in cases of AI-related errors or misuse poses a significant challenge.

Common Obstacles

• Limited access to diverse and unbiased data sets hampers the development of fair AI models.
• Gaining public confidence in AI-assisted policing is hindered by concerns about surveillance and misuse of personal data.
• Limited resources hinder the deployment of advanced AI systems, affecting both training and implementation.

Info-Tech's Approach

• Info-Tech's guidance provides for meticulous data curation, transparency, and ongoing bias mitigation efforts in responsible AI model development.
• Our responsible AI implementation strategy:
• Identifies core responsible AI principles as sources of value to strategically address challenges and safely, securely, and fairly implement initiatives;
• Jumpstarts the idea generation process during the initiative development phase;
• Offers six insights for responsible use of AI in policing;
• Provides next steps toward AI-driven initiative integration and implementation; and
• Builds in safeguards to foster public trust and community engagement.

Info-Tech Insight

By ensuring the responsible and ethical use of AI in policing, and getting the public involved in AI in policing development, law enforcement agencies can harness its potential, while minimizing its pitfalls, and ultimately, enhance the effectiveness, efficiency, and accountability of law enforcement agencies, and the safety, security, and wellbeing of the community being served.

Section 1

Six Key Insights for the Responsible Use of AI in Policing

AI in policing poses unique risks to the public

Insight No. 1

Ethical, legal, and social implications

AI in policing may raise issues of privacy, consent, fairness, accountability, and oversight, as it may collect, store, share, and use sensitive and personal data, without the knowledge or consent of the data subjects, and may affect their lives and opportunities in significant ways.

Potential biases and errors

AI in policing may introduce or amplify biases and errors, as it may reflect or reproduce the existing inequalities, prejudices, and stereotypes in the data, algorithms, or systems, and may generate inaccurate or unreliable results or recommendations.

Public trust and acceptance

AI in policing may affect the public trust and acceptance of law enforcement agencies, as it may create or increase the perception of surveillance, intrusion, manipulation, or discrimination, and may undermine the human dignity and autonomy of the individuals and communities.

Address the risks to avoid harming the public

Insight No. 2

Ethical principles and guidelines

Developing and applying ethical principles and guidelines for AI in policing that are aligned with universal human rights and values, and that address the specific challenges and needs of the field.

Compliance and accountability

Implementing and monitoring the compliance and accountability mechanisms for AI in policing that ensure the legality, quality, and validity of the data, algorithms, and systems, and that provide the means and avenues for oversight, audit, review, and redress.

Education and awareness

Promoting and supporting the education and awareness of AI in policing that inform and equip the law enforcement personnel, the public, and other stakeholders, with the necessary knowledge, skills, and competencies to understand, use, and evaluate AI in policing.

Participation and collaboration

Fostering and facilitating the participation and collaboration of AI in policing that involve and consult law enforcement personnel, the public, and other stakeholders in the design, development, deployment, and evaluation of AI in policing, and that respect and balance their interests, needs, and expectations.

Employ ethical principles and guidelines for AI in policing

Insight No. 3

Responsible

AI in policing should be used for lawful, legitimate, and appropriate purposes, and should respect and protect the human dignity, rights, and values of all parties involved.

Equitable

AI in policing should be fair, impartial, and non-discriminatory, and should avoid or mitigate any potential biases, errors, or harms that may arise from the data, algorithms, or systems.

Traceable

AI in policing should be transparent, explainable, and accountable, and should provide clear and accessible information about the data, algorithms, and systems, as well as their sources, methods, outcomes, and impacts.

Reliable

AI in policing should be accurate, consistent, and robust, and should ensure the quality, validity, and security of the data, algorithms, and systems, as well as their performance, functionality, and reliability.

Governable

AI in policing should be controllable, adaptable, and responsive, and should provide the means and mechanisms for oversight, audit, review, and redress, as well as for human intervention and override.

Download Info-Tech's Develop Responsible AI Guiding Principles

Pursue best practices that follow strict ethical principles

Insight No. 4

Automatic patrol systems

These are AI systems that use drones or robots to patrol certain areas and detect any suspicious or criminal activities, such as vandalism, theft, or violence. They can also alert the human police officers and provide them with real-time information and evidence. These systems can help improve the safety and efficiency of law enforcement, while respecting the privacy and rights of the citizens.

Identification of vulnerable and exploited children

These are AI systems that use facial recognition and biometrics to identify and rescue children who are victims of human trafficking, sexual exploitation, or other forms of abuse. They can also help locate and prosecute the perpetrators and provide support and protection to the children. These systems can help prevent and reduce the harm and suffering of the children, while ensuring their dignity and wellbeing.

Police emergency call centers

These are AI systems that use natural language processing and speech recognition to handle and prioritize the calls from the public who need police assistance. They can also provide the callers with relevant and timely information, guidance, and feedback, and connect them with the human police officers if needed. These systems can help enhance the communication and collaboration between the police and the public, while ensuring the quality and reliability of the service.

Ensure the equitable use of AI in policing to build public trust

Insight No. 5

Equitable use of AI in policing means that AI is used in a fair, impartial, and non-discriminatory way, and that it avoids or mitigates any potential biases, errors, or harms that may arise from the data, algorithms, or systems.

Collecting and using the right data

The data used to train and test AI systems should be representative, relevant, and reliable, and should not contain any biases, inaccuracies, or gaps that may affect the outcomes or impacts of AI in policing.

Tailoring contracting approaches

The procurement and contracting of AI systems should be transparent, competitive, and accountable, and should specify the requirements, expectations, and responsibilities of the parties involved, as well as the performance, functionality, and reliability of the AI systems.

Developing governance structures

The governance of AI systems should ensure the oversight, audit, review, and redress of the data, algorithms, and systems, as well as the human intervention and override, and should provide clear and accessible information and communication to law enforcement personnel, the public, and other stakeholders.