ABOUT MODULE 3
MODULE 3
STRUCTURE & OBJECTIVES
Click
Section 1
Click
Section 2
Section 3
MODULE 3
REMINDER
Click
Click to reveal a checklist of practical tips to succeed in your online training.
Go
SECTION 1
UNDERSTANDING THE PRINCIPLES
SECTION 2
THE PRINCIPLES OF RESPONSIBLE AI INNOVATION
The principles – both core and instrumental – are explained in more detail in the following slides. Moreover, the principles are described in such a manner that allows them to be adapted to the diverse contexts in which law enforcement operates. Start by clicking on the icons to listen to a brief introduction of each one of the core principles.
Go
CORE PRINCIPLE 1: LAWFULNESS
01:02
Proportionality
Legitimacy
Necessity
Instrumental principles
CORE PRINCIPLE 2: MINIMIZATION OF HARM
01:19
Human and Environmental Well-being
Robustness and Safety
Accuracy
Efficiency
Instrumental principles
CORE PRINCIPLE 3: HUMAN AUTONOMY
00:37
Privacy
Human Control and Oversight
Human Agency
Transparency and Explainability
Instrumental principles
CORE PRINCIPLE 4: FAIRNESS
01:06
Diversity and Accessibility
Equality and Non-discrimination
Protection of Vulnerable Groups
Contestability and Redress
Instrumental principles
CORE PRINCIPLE 5: GOOD GOVERNANCE
00:50
Accountability
Traceability and Auditability
Instrumental principles
SECTION 3
PUTTING THE PRINCIPLES INTO PRACTICE
How to put the principles in practice?
Let's see in a nutshell what these steps entail.
Go
Congratulations!!! You have reached the end of Module 3
The principle of human and environmental well-being entails law enforcement agencies preserving and improving the welfare of people and the environment in their AI innovation journey. This consideration is partially ensured by the principles of robustness, safety, and accuracy. However, human and environmental well-being is a broader principle, as it implies that agencies should examine the full spectrum of direct and indirect consequences of their AI-related activities and aim for the improvement of well-being. By combining societal and environmental sustainability issues, this principle can facilitate discussion and consideration of matters such as energy consumption and the use of resources during all phases of the AI system’s life cycle. In this sense, it is also connected with the principle of efficiency.
Good governance in AI innovation in law enforcement calls for agencies to set up requirements, procedures, and technical solutions to ensure that the decision-making processes of an AI system are traceable, including adequately documenting the decisions made during design, development and use that influence the outputs of the AI system. During use, traceability involves tracking and documenting AI outputs, including the input data used, the model and parameters selected, the model’s output, the user’s name, date, and any other relevant information. In addition, law enforcement agencies should ensure that the AI systems they use are auditable, in that their essential elements can be assessed by internal or external auditors.
The Principles for responsible AI innovation
The learning objective of this section is to be able to describe each one of the core principles and unpack their associated instrumental principles.
In the context of AI, human control and oversight are the ability and opportunity for humans to adequately supervise, engage and interfere with an AI system during its development and use. To ensure human control and oversight, law enforcement agencies are advised to verify that the AI systems they currently use or intend to use are built with the functionalities needed to ensure that humans remain in charge during use, as well as to confirm that the necessary organizational structures are in place to ensure that humans have the last word regarding certain decisions. Upholding human control and oversight of AI systems is particularly important in the law enforcement context. This is especially true considering that the work of law enforcement agencies is at the very core of the functioning of society, justice, and political systems, and therefore has a significant influence on individuals and their rights.
In the context of AI innovation in law enforcement, diversity and accessibility mean that AI systems should be built to be usable by a wide range of individuals and groups, regardless of age, gender, ability, or other characteristics. This means verifying that the systems that are developed, procured and deployed are designed in a user-centric way and account for the various characteristics and abilities that the end users may have. Building inclusive systems is crucial whenever these systems have an impact on people accessing public goods, services, or advantages. This principle is thus particularly relevant when law enforcement agencies develop, procure or use AI systems that are intended for use by the general public, as diversity and accessibility in the systems’ design will have a direct impact on societal fairness. In fact, like any other tool, AI systems can empower people, or they can disenfranchise them due to lack of accessibility.
Necessity means that law enforcement agencies should only interfere with people’s rights when such interferences are needed to fulfil the identified legitimate goal. This means that, even when the pursued goal is legitimate, agencies should ensure that it cannot be achieved without interfering with human rights. They should also note that while interference may be necessary at first, it may become unnecessary if the goal is achieved or can no longer be achieved in a lawful way.
When AI systems are used for decision-making processes in a law enforcement setting, it is crucial that mechanisms are put in place to enable stakeholders to clearly determine who is responsible for the decisions made with the support of the AI system, and the consequences of those decisions. The central role that accountability plays in this context relates to the prominence of law enforcement in the functioning of society, justice, and governments, and consequently the high stakes for everyone involved. Because of the authority accorded to law enforcement agencies and officers, which is essential for the pursuit of their mission, there is an inherent power imbalance between those in charge of law enforcement and the rest of society. The complexity of AI systems, combined with the general population’s lack of understanding of AI, could exacerbate this power imbalance when these systems are introduced. Responsible AI innovation compensates for this imbalance by requiring processes to be put in place to clearly determine which individuals are accountable for AI-related decisions.
Accuracy corresponds to the degree to which an AI system can make correct predictions, recommendations, or decisions. It is important that agencies verify that any system they are developing and/or intend to use is highly accurate, as using inaccurate AI systems can result in various types of harm. For example, if an AI system used for crime detection has a low accuracy rate, it could potentially cause law enforcement officers to be misled into responding to a location where no actual crime has occurred. This could be detrimental to both law enforcement agencies and society as a whole, as it would result in the unnecessary waste of valuable and often scarce resources. Therefore, before deploying an AI system into mainstream application in the law enforcement context, such system needs to be subject to rigorous and scientific testing.
Efficiency in AI innovation means that law enforcement agencies make sure that there is a favorable ratio between the costs and the benefits of using a certain AI system in terms of time, money, human effort, and the impact on the environment. One of AI’s biggest promises is efficiency. Using AI systems can allow complex tasks to be completed in a faster, easier, and less-resource intensive manner. However, costs are incurred at all stages of the AI system’s life cycle. For example, agencies need to spend money, time, and human and environmental resources on developing, procuring, and deploying a good system, including training personnel to use and monitor it, and purchasing adequate hardware for it to run. The efficiency principle requires agencies to determine whether the benefits of using the system outweigh the costs.
6. Take Timed Breaks
5. Take notes
4. Keep Away from Social Media
To pursue fair AI innovation, law enforcement agencies should pay particular attention and due consideration to those groups who are most vulnerable to and at risk of being disadvantaged by the use of specific AI systems. Safeguards should be put in place throughout the AI life cycle to mitigate the risks and enhance the benefits for these groups. Through their design, development, deployment and use, AI systems may have a disproportionately negative impact on certain groups due to their characteristics or other circumstances. For example, differences in the accuracy of AI systems often affect certain groups more than others, especially because certain groups are more susceptible to being misrepresented in the data sets that are used to train the systems.
To safeguard human autonomy in the context of AI innovation in law enforcement, it is important that agencies engage with AI systems in a way that protects the private sphere of individuals, including the users of the AI system, victims, suspects, and the general public. This entails safeguarding their physical and mental integrity, personal relationships, personal space and home, and personal data in general, as this is essential for individuals to maintain their capacity to self-govern and exercise their rights. Respecting privacy is a general condition of principled policing. By its very nature, law enforcement work requires the collection and analysis of information often related to the private lives of individuals. Therefore, the duty of confidentiality is a common element across the various professional rules for law enforcement officers.
Respecting equality and non-discrimination within AI innovation in law enforcement means ensuring equal treatment and opportunities for all stakeholders and refraining from unjustifiably discriminating against individuals or groups throughout the AI life cycle. Equality and non-discrimination are especially important in the context of responsible AI innovation in law enforcement. To cultivate responsible AI, law enforcement agencies need to ensure that the AI systems they use are trained with data sets containing the appropriate quality and quantity of data and that any identifiable and discriminatory biases are removed. Any decisions taken in the design and development of the system that may have a negative, unfair, or disproportionate impact on certain individuals or groups also need to be considered.
1. Make sure your computer and Internet connection work properly
There is an increased expectation from workers, criminal justice practitioners, regulators, and society in general that they will be involved in high stakes decisions related to AI innovation in law enforcement. Successfully implementing new AI systems in an agency therefore requires identifying and engaging with the relevant stakeholders. In the context of law enforcement, these stakeholders may include:
- The individuals who are subject to and may benefit from or be harmed by the use of an AI system, such as suspects, victims, civil society groups, and the general public.
- The individuals whose data is used to test and develop AI systems.
- Innovation units and development teams both within law enforcement and in the private sector who develop AI systems and tools.
- Law enforcement officers and other personnel who interact with AI systems.
- Law enforcement management, who will be accountable for deploying an AI system too early or for missing an opportunity to use an AI system.
- Practitioners within criminal justice systems who need to make sense of the information and decisions that they receive from law enforcement.
The principle of contestability means that law enforcement agencies should ensure that the necessary technological and organizational measures are in place to allow both users and those affected by decisions based on the output of an AI system to challenge these decisions. Contestability focuses on the ability to argue against AI-supported decisions. It is linked to human control and oversight, transparency and explainability as well as good governance and its instrumental principles, in that all these principles are requisites to properly fulfilling the principle of contestability. The principle of redress means that agencies should go one step further and ensure that, when AI-supported decisions have an unjust negative impact, those affected are able to formally seek redress through adequate and accessible processes. Upholding the principle of redress also relates to the human right to an effective remedy, and therefore to the principle of lawfulness.
Putting the Principles into practice
The learning objective of this section is to be able to follow the process allowing to put the Principles for responsible AI innovation into practice.
Responsible AI innovation entails that the people that interact with AI systems have enough knowledge and understanding of the systems to safeguard their autonomy. This is especially relevant in a law enforcement setting given the nature and the impact of law enforcement work and can be achieved by following the principles of transparency and explainability. These are related but distinct principles: while transparency focuses on promoting good communication practices throughout the AI life cycle, explainability aims to allow individuals to understand how the system reaches its outcomes. To ensure transparency, law enforcement agencies are advised to verify that the developers of their AI system (internal or external) disclose all the necessary information and documentation to its users.
Understanding the principles
The learning objective of this section is to be able to identify the five core principles of responsible AI innovation and the complementary function of the instrumental principles.
7. DO NOT Multitask
As law enforcement agencies advance through the AI life cycle, they should keep in mind the principles and the relevant stakeholders. Agencies are advised to keep track of the consequences of their decisions and the results of their activities, and correct their course if needed.
2. Enable sound on your computer and connect a headphone set
Human agency is the ability of a person to act upon their own decisions and pursue their goals without manipulation or force. To protect human agency in the context of AI innovation, law enforcement agencies need to ensure that the AI systems they aim to use do not compromise the ability of the users of those systems (law enforcement officers, other personnel, citizens, etc.) to act and make decisions independently. Human agency can be challenged if individuals or institutions are over reliant on AI systems, disregarding human input when it may be relevant or even necessary. For example, if an AI system is used in a certain process, agencies should in most cases ensure that the system is genuinely supporting or improving the decisions taken by the officers in charge of the process, instead of making those decisions for them. This also entails training the officers, so they know how to engage properly with the AI system.
The principles are meant to be followed throughout the AI life cycle to support all decision-makers in a law enforcement agency in evaluating the impact of an AI system on individuals, society, and the environment, and establishing the measures that can be taken to avoid or mitigate any negative consequences.
In practice, this involves asking different questions at each stage, thus allowing agencies to thoroughly explore and address the positive and negative consequences of implementing any given AI system.
Law enforcement officers in their various relevant capacities are recommended to have a good understanding of the principles from the beginning of their engagement with an AI system.
Legitimacy means that law enforcement agencies should only interfere with people’s rights when they have a valid reason to do so, based on domestic law and in line with international standards. This means that, for any interference with human rights, law enforcement agencies need to fulfil two requirements from the beginning:
- Having a legal basis for that interference; and
- Following a legitimate goal such as safeguarding the life and safety of individuals and society.
3. Put electronic gadgets away or on silence mode
Proportionality means that law enforcement agencies need to balance the interference with human rights against the reason for doing it (the legitimate goal). This implies that interferences must always correspond to the least intrusive way of achieving such a goal and that the negative effects they have on people’s rights must be balanced against the legitimate goal pursued. This balancing exercise is also closely connected with the core principle of fairness.
Robustness and safety imply that AI systems can maintain consistency across different contexts and identify and prevent potential risks of harm, and that they are protected against attacks and overall do not pose a threat to the physical or mental well-being of individuals, their property, or the environment. Given the central role that robustness and safety play in preventing and minimizing the risks of harm posed by the use of AI systems, responsible AI innovation in law enforcement requires agencies to verify that the AI systems they are developing and using are built in line with these principles. More specifically, to ensure robustness, law enforcement agencies should confirm that the AI systems they intend to use are both reliable and secure.
REVIEW (FOR EXTERNAL SHARING) - MODULE 3
alejandro.rojas.ramo
Created on February 23, 2026
Start designing with a free template
Discover more than 1500 professional designs like these:
View
Smart Presentation
View
Practical Presentation
View
Essential Presentation
View
Akihabara Presentation
View
Flow Presentation
View
Terrazzo Presentation
View
Dynamic Visual Presentation
Explore all templates
Transcript
ABOUT MODULE 3
MODULE 3
STRUCTURE & OBJECTIVES
Click
Section 1
Click
Section 2
Section 3
MODULE 3
REMINDER
Click
Click to reveal a checklist of practical tips to succeed in your online training.
Go
SECTION 1
UNDERSTANDING THE PRINCIPLES
SECTION 2
THE PRINCIPLES OF RESPONSIBLE AI INNOVATION
The principles – both core and instrumental – are explained in more detail in the following slides. Moreover, the principles are described in such a manner that allows them to be adapted to the diverse contexts in which law enforcement operates. Start by clicking on the icons to listen to a brief introduction of each one of the core principles.
Go
CORE PRINCIPLE 1: LAWFULNESS
01:02
Proportionality
Legitimacy
Necessity
Instrumental principles
CORE PRINCIPLE 2: MINIMIZATION OF HARM
01:19
Human and Environmental Well-being
Robustness and Safety
Accuracy
Efficiency
Instrumental principles
CORE PRINCIPLE 3: HUMAN AUTONOMY
00:37
Privacy
Human Control and Oversight
Human Agency
Transparency and Explainability
Instrumental principles
CORE PRINCIPLE 4: FAIRNESS
01:06
Diversity and Accessibility
Equality and Non-discrimination
Protection of Vulnerable Groups
Contestability and Redress
Instrumental principles
CORE PRINCIPLE 5: GOOD GOVERNANCE
00:50
Accountability
Traceability and Auditability
Instrumental principles
SECTION 3
PUTTING THE PRINCIPLES INTO PRACTICE
How to put the principles in practice?
Let's see in a nutshell what these steps entail.
Go
Congratulations!!! You have reached the end of Module 3
The principle of human and environmental well-being entails law enforcement agencies preserving and improving the welfare of people and the environment in their AI innovation journey. This consideration is partially ensured by the principles of robustness, safety, and accuracy. However, human and environmental well-being is a broader principle, as it implies that agencies should examine the full spectrum of direct and indirect consequences of their AI-related activities and aim for the improvement of well-being. By combining societal and environmental sustainability issues, this principle can facilitate discussion and consideration of matters such as energy consumption and the use of resources during all phases of the AI system’s life cycle. In this sense, it is also connected with the principle of efficiency.
Good governance in AI innovation in law enforcement calls for agencies to set up requirements, procedures, and technical solutions to ensure that the decision-making processes of an AI system are traceable, including adequately documenting the decisions made during design, development and use that influence the outputs of the AI system. During use, traceability involves tracking and documenting AI outputs, including the input data used, the model and parameters selected, the model’s output, the user’s name, date, and any other relevant information. In addition, law enforcement agencies should ensure that the AI systems they use are auditable, in that their essential elements can be assessed by internal or external auditors.
The Principles for responsible AI innovation
The learning objective of this section is to be able to describe each one of the core principles and unpack their associated instrumental principles.
In the context of AI, human control and oversight are the ability and opportunity for humans to adequately supervise, engage and interfere with an AI system during its development and use. To ensure human control and oversight, law enforcement agencies are advised to verify that the AI systems they currently use or intend to use are built with the functionalities needed to ensure that humans remain in charge during use, as well as to confirm that the necessary organizational structures are in place to ensure that humans have the last word regarding certain decisions. Upholding human control and oversight of AI systems is particularly important in the law enforcement context. This is especially true considering that the work of law enforcement agencies is at the very core of the functioning of society, justice, and political systems, and therefore has a significant influence on individuals and their rights.
In the context of AI innovation in law enforcement, diversity and accessibility mean that AI systems should be built to be usable by a wide range of individuals and groups, regardless of age, gender, ability, or other characteristics. This means verifying that the systems that are developed, procured and deployed are designed in a user-centric way and account for the various characteristics and abilities that the end users may have. Building inclusive systems is crucial whenever these systems have an impact on people accessing public goods, services, or advantages. This principle is thus particularly relevant when law enforcement agencies develop, procure or use AI systems that are intended for use by the general public, as diversity and accessibility in the systems’ design will have a direct impact on societal fairness. In fact, like any other tool, AI systems can empower people, or they can disenfranchise them due to lack of accessibility.
Necessity means that law enforcement agencies should only interfere with people’s rights when such interferences are needed to fulfil the identified legitimate goal. This means that, even when the pursued goal is legitimate, agencies should ensure that it cannot be achieved without interfering with human rights. They should also note that while interference may be necessary at first, it may become unnecessary if the goal is achieved or can no longer be achieved in a lawful way.
When AI systems are used for decision-making processes in a law enforcement setting, it is crucial that mechanisms are put in place to enable stakeholders to clearly determine who is responsible for the decisions made with the support of the AI system, and the consequences of those decisions. The central role that accountability plays in this context relates to the prominence of law enforcement in the functioning of society, justice, and governments, and consequently the high stakes for everyone involved. Because of the authority accorded to law enforcement agencies and officers, which is essential for the pursuit of their mission, there is an inherent power imbalance between those in charge of law enforcement and the rest of society. The complexity of AI systems, combined with the general population’s lack of understanding of AI, could exacerbate this power imbalance when these systems are introduced. Responsible AI innovation compensates for this imbalance by requiring processes to be put in place to clearly determine which individuals are accountable for AI-related decisions.
Accuracy corresponds to the degree to which an AI system can make correct predictions, recommendations, or decisions. It is important that agencies verify that any system they are developing and/or intend to use is highly accurate, as using inaccurate AI systems can result in various types of harm. For example, if an AI system used for crime detection has a low accuracy rate, it could potentially cause law enforcement officers to be misled into responding to a location where no actual crime has occurred. This could be detrimental to both law enforcement agencies and society as a whole, as it would result in the unnecessary waste of valuable and often scarce resources. Therefore, before deploying an AI system into mainstream application in the law enforcement context, such system needs to be subject to rigorous and scientific testing.
Efficiency in AI innovation means that law enforcement agencies make sure that there is a favorable ratio between the costs and the benefits of using a certain AI system in terms of time, money, human effort, and the impact on the environment. One of AI’s biggest promises is efficiency. Using AI systems can allow complex tasks to be completed in a faster, easier, and less-resource intensive manner. However, costs are incurred at all stages of the AI system’s life cycle. For example, agencies need to spend money, time, and human and environmental resources on developing, procuring, and deploying a good system, including training personnel to use and monitor it, and purchasing adequate hardware for it to run. The efficiency principle requires agencies to determine whether the benefits of using the system outweigh the costs.
6. Take Timed Breaks
5. Take notes
4. Keep Away from Social Media
To pursue fair AI innovation, law enforcement agencies should pay particular attention and due consideration to those groups who are most vulnerable to and at risk of being disadvantaged by the use of specific AI systems. Safeguards should be put in place throughout the AI life cycle to mitigate the risks and enhance the benefits for these groups. Through their design, development, deployment and use, AI systems may have a disproportionately negative impact on certain groups due to their characteristics or other circumstances. For example, differences in the accuracy of AI systems often affect certain groups more than others, especially because certain groups are more susceptible to being misrepresented in the data sets that are used to train the systems.
To safeguard human autonomy in the context of AI innovation in law enforcement, it is important that agencies engage with AI systems in a way that protects the private sphere of individuals, including the users of the AI system, victims, suspects, and the general public. This entails safeguarding their physical and mental integrity, personal relationships, personal space and home, and personal data in general, as this is essential for individuals to maintain their capacity to self-govern and exercise their rights. Respecting privacy is a general condition of principled policing. By its very nature, law enforcement work requires the collection and analysis of information often related to the private lives of individuals. Therefore, the duty of confidentiality is a common element across the various professional rules for law enforcement officers.
Respecting equality and non-discrimination within AI innovation in law enforcement means ensuring equal treatment and opportunities for all stakeholders and refraining from unjustifiably discriminating against individuals or groups throughout the AI life cycle. Equality and non-discrimination are especially important in the context of responsible AI innovation in law enforcement. To cultivate responsible AI, law enforcement agencies need to ensure that the AI systems they use are trained with data sets containing the appropriate quality and quantity of data and that any identifiable and discriminatory biases are removed. Any decisions taken in the design and development of the system that may have a negative, unfair, or disproportionate impact on certain individuals or groups also need to be considered.
1. Make sure your computer and Internet connection work properly
There is an increased expectation from workers, criminal justice practitioners, regulators, and society in general that they will be involved in high stakes decisions related to AI innovation in law enforcement. Successfully implementing new AI systems in an agency therefore requires identifying and engaging with the relevant stakeholders. In the context of law enforcement, these stakeholders may include:
The principle of contestability means that law enforcement agencies should ensure that the necessary technological and organizational measures are in place to allow both users and those affected by decisions based on the output of an AI system to challenge these decisions. Contestability focuses on the ability to argue against AI-supported decisions. It is linked to human control and oversight, transparency and explainability as well as good governance and its instrumental principles, in that all these principles are requisites to properly fulfilling the principle of contestability. The principle of redress means that agencies should go one step further and ensure that, when AI-supported decisions have an unjust negative impact, those affected are able to formally seek redress through adequate and accessible processes. Upholding the principle of redress also relates to the human right to an effective remedy, and therefore to the principle of lawfulness.
Putting the Principles into practice
The learning objective of this section is to be able to follow the process allowing to put the Principles for responsible AI innovation into practice.
Responsible AI innovation entails that the people that interact with AI systems have enough knowledge and understanding of the systems to safeguard their autonomy. This is especially relevant in a law enforcement setting given the nature and the impact of law enforcement work and can be achieved by following the principles of transparency and explainability. These are related but distinct principles: while transparency focuses on promoting good communication practices throughout the AI life cycle, explainability aims to allow individuals to understand how the system reaches its outcomes. To ensure transparency, law enforcement agencies are advised to verify that the developers of their AI system (internal or external) disclose all the necessary information and documentation to its users.
Understanding the principles
The learning objective of this section is to be able to identify the five core principles of responsible AI innovation and the complementary function of the instrumental principles.
7. DO NOT Multitask
As law enforcement agencies advance through the AI life cycle, they should keep in mind the principles and the relevant stakeholders. Agencies are advised to keep track of the consequences of their decisions and the results of their activities, and correct their course if needed.
2. Enable sound on your computer and connect a headphone set
Human agency is the ability of a person to act upon their own decisions and pursue their goals without manipulation or force. To protect human agency in the context of AI innovation, law enforcement agencies need to ensure that the AI systems they aim to use do not compromise the ability of the users of those systems (law enforcement officers, other personnel, citizens, etc.) to act and make decisions independently. Human agency can be challenged if individuals or institutions are over reliant on AI systems, disregarding human input when it may be relevant or even necessary. For example, if an AI system is used in a certain process, agencies should in most cases ensure that the system is genuinely supporting or improving the decisions taken by the officers in charge of the process, instead of making those decisions for them. This also entails training the officers, so they know how to engage properly with the AI system.
The principles are meant to be followed throughout the AI life cycle to support all decision-makers in a law enforcement agency in evaluating the impact of an AI system on individuals, society, and the environment, and establishing the measures that can be taken to avoid or mitigate any negative consequences.
In practice, this involves asking different questions at each stage, thus allowing agencies to thoroughly explore and address the positive and negative consequences of implementing any given AI system.
Law enforcement officers in their various relevant capacities are recommended to have a good understanding of the principles from the beginning of their engagement with an AI system.
Legitimacy means that law enforcement agencies should only interfere with people’s rights when they have a valid reason to do so, based on domestic law and in line with international standards. This means that, for any interference with human rights, law enforcement agencies need to fulfil two requirements from the beginning:
3. Put electronic gadgets away or on silence mode
Proportionality means that law enforcement agencies need to balance the interference with human rights against the reason for doing it (the legitimate goal). This implies that interferences must always correspond to the least intrusive way of achieving such a goal and that the negative effects they have on people’s rights must be balanced against the legitimate goal pursued. This balancing exercise is also closely connected with the core principle of fairness.
Robustness and safety imply that AI systems can maintain consistency across different contexts and identify and prevent potential risks of harm, and that they are protected against attacks and overall do not pose a threat to the physical or mental well-being of individuals, their property, or the environment. Given the central role that robustness and safety play in preventing and minimizing the risks of harm posed by the use of AI systems, responsible AI innovation in law enforcement requires agencies to verify that the AI systems they are developing and using are built in line with these principles. More specifically, to ensure robustness, law enforcement agencies should confirm that the AI systems they intend to use are both reliable and secure.