README

File

REVISED FEBRUARY 2024

I
DISCLAIMER

The contents of this document are for information purposes only. INTERPOL and UNICRI assume
no liability or responsibility for any inaccurate or incomplete information, nor for any actions
taken in reliance thereon. The published material is distributed without warranty of any kind,
either express or implied, and the responsibility for the interpretation and use of the material lies
with the reader. In no event shall, INTERPOL or UNICRI be liable for damages arising from its use.

INTERPOL and UNICRI take no responsibility for the content of any external website referenced
in this publication or for any defamatory, offensive or misleading information which might be
contained on these third-party websites. Any links to external websites do not constitute an
endorsement by INTERPOL or UNICRI, and are only provided as a convenience. It is the responsibility
of the reader to evaluate the content and usefulness of information obtained from other sites.

The views, thoughts and opinions expressed in the content of this publication belong solely to the
authors and do not necessarily reflect the views or policies of, nor do they imply any endorsement
by, INTERPOL or the United Nations, their member countries or member states, their governing
bodies, or contributory organizations. Therefore, INTERPOL and UNICRI carry no responsibility for
the opinions expressed in this publication.

INTERPOL and UNICRI do not endorse or recommend any product, process, or service. Therefore,
mention of any products, processes, or services in this document cannot be construed as an
endorsement or recommendation by INTERPOL or UNICRI.

The designation employed and presentation of the material in this document do not imply the
expression of any opinion whatsoever on the part of the Secretariat of the United Nations, UNICRI
or INTERPOL, concerning the legal status of any country, territory, city or area of its authorities, or
concerning the delimitation of its frontiers or boundaries.

The contents of this document may be quoted or reproduced, provided that the source of
information is acknowledged. INTERPOL and UNICRI would like to receive a copy of the document
in which this publication is used or quoted.
 FOREWORD

In the continuously evolving realm of technology, Artificial Intelligence

(AI) has emerged as a formidable force capable of reshaping almost
every facet of society.

In recent years, we have witnessed incredible progress in AI technologies

and its potential as an asset, but also as a threat. AI’s immense power
and potential benefits require a strong sense of responsibility, especially
for law enforcement.

As our society embraces these technological advancements, it becomes

paramount to ensure that the deployment of AI in policing is guided by an
unwavering commitment to upholding ethical standards, safeguarding
human rights, and fostering fairness and transparency.

Recognizing the concerns expressed by representatives from the global

law enforcement community regarding the lack of guidance in utilizing
AI, INTERPOL and the United Nations Interregional Crime and Justice
Research Institute (UNICRI) have jointly developed this Toolkit for
Responsible AI Innovation in Law Enforcement.

The Toolkit aims to provide support and guidance to law enforcement

agencies, enabling them to leverage AI in a trustworthy, lawful and
responsible manner, providing a clear, pragmatic, and most of all useful
guide.

As a unique global platform for law enforcement cooperation, INTERPOL

is committed to making both the physical and virtual realms safer. My
thanks to everyone who contributed to the development of this AI Toolkit
which exemplifies the collaborative efforts required to navigate the
intricate landscape of AI in law enforcement.
I particularly extend my gratitude and appreciation to UNICRI for their
invaluable collaboration, unwavering support, and commitment to our
shared goals. We look forward to continuing this fruitful partnership.

INTERPOL’s databases are at the core of the services we offer member

countries, so any way tools can help connect police with the right
intelligence more efficiently merits being explored.

Together with the recently established INTERPOL Responsible AI Lab,

the Toolkit stands as a testament to INTERPOL’s ongoing commitment
to fostering an innovative and proactive law enforcement culture.

As INTERPOL celebrate its centenary, the AI Toolkit also highlights the

Organization’s flexibility whilst remaining true to our founding principles
as we coordinate global law enforcement efforts addressing the
increasingly complex crime threat landscape of today, and tomorrow.

JUERGEN STOCK
Secretary General, INTERPOL
 FOREWORD

We find ourselves in extraordinary times. Artificial Intelligence (AI) is

visibly changing the dynamics of the world around us. It is exhilarating,
inspiring, and promising, and yet, it also demands our utmost vigilance
and responsibility – particularly in the law enforcement context, which is
highly sensitive.

UNICRI is working at the forefront of this technological revolution. As

the United Nations research and training institute with a focus on crime
prevention and criminal justice, we are heavily invested in responsible
AI and ‘getting it right’. This flows from the commitment in the United
Nations Charter to promote and encourage respect for human rights and
fundamental freedoms for all, without discrimination. A key example of
this commitment is UNESCO’s groundbreaking Recommendation on the
Ethics of Artificial Intelligence, a unique framework adopted by all 193
Member States in 2021. At UNICRI, we have worked tirelessly in this
direction since the establishment of our Centre for AI and Robotics in
2017. This Toolkit for Responsible AI Innovation in Law Enforcement is
our contribution to shaping the trajectory of responsible AI innovation in
law enforcement.

We are proud to partner with INTERPOL on this crucial endeavour.

INTERPOL has been the lynchpin of international police cooperation for
100 years, and it has been a key ally of the United Nations system since
its own founding in 1945. INTERPOL is a trusted partner of UNICRI for
many years already, enriching the depth and breadth of our joint work.
Together we have produced an invaluable blueprint to guide the global
law enforcement community to leverage the promise of AI in a human
rights compliant and ethical manner. This has been our shared vision
and mission since day one of our joint work on responsible AI innovation.
Finally, this endeavour has benefitted greatly from the commitment and
support of the European Union. Through the General Data Protection
Regulation (GDPR), the European Commission’s High-Level Expert
Group on Artificial Intelligence, the European AI Alliance, and the
proposed European Artificial Intelligence Act, Europe has demonstrated
its commitment to responsible AI, recognizing the need to balance
innovation with ethics, human rights, and societal well-being. We now
add one more example to that impressive list: the Toolkit for Responsible
AI Innovation in Law Enforcement.

As we stand at the precipice of this technological revolution, we bear

witness to the immense opportunities and challenges of AI. By embracing
the principles and guidance outlined in this AI Toolkit, we can harness the
potential of AI in law enforcement to help improve public safety, prevent
crime, and promote justice for victims.

Together, let us strive towards a future where innovation and human

rights go hand in hand, taking advantage of the benefits and mitigating
the risks of AI, to build a safer and more equitable world for all.

ANTONIA MARIA DE MEO

Director, UNICRI
 README File

What is the AI Toolkit?

Artificial intelligence (AI) is transforming the world, and policing is no exception to this: law
enforcement agencies in many parts of the world are already developing, acquiring and using
different types of AI systems, tools, and technologies to support an increasingly broad range of
activities. Whether for image, text and speech analysis or risk assessment, AI is already essential
for many agencies, and it is on its way to becoming one of the fundamental operating tools of
policing.

The Toolkit for Responsible AI Innovation in Law Enforcement is a tailor-made set of resources for
the global law enforcement community, developed by INTERPOL and UNICRI with the financial
support of the European Union. Its objective is to support law enforcement agencies to navigate
the complex task of institutionalizing responsible AI innovation and integrating AI systems into
their work. It has been designed in full recognition of both the opportunities and the challenges
that AI presents for the law enforcement community, and it is intended to offer support to agencies
interested in or already committed to responsible AI innovation. More specifically, this AI Toolkit
provides law enforcement agencies with a theoretical foundation on responsible AI innovation
based on human rights law, ethics and policing principles, as well as several practical tools to
support them with putting responsible AI innovation theory into practice at each and every stage
of their AI journey.

6
 README File

This AI Toolkit is comprised of seven individual standalone resources. Together, these resources
provide agencies with the necessary knowledge and understanding to be able to benefit from the
positive potential of these systems while navigating the possible challenges and risks, whatever
the agency’s level of experience with AI. The resources have all been designed to be informative,
practical and actionable, providing guidance and best practices to move forward along the path
of responsible AI innovation.

The seven resources are grouped into three categories (see the table):

PRIMARY GUIDANCE SUPPORTING

PRACTICAL TOOLS
DOCUMENTS DOCUMENTS

Introduction to Organizational Technical Reference

Responsible AI Readiness Assessment Book
Innovation Questionnaire

Principles for Risk Assessment

Responsible AI Questionnaire
Innovation
Responsible AI
Organizational Innovation in Action
Roadmap Workbook

The AI Toolkit is designed to provide the law enforcement community with a balanced perspective
on responsible AI innovation that is grounded in human rights, ethics, and established good
policing principles, all contextualized with practical examples of law enforcement-specific use
cases which are provided throughout the AI Toolkit.

It is important to note that the AI Toolkit is not intended to create binding obligations or to limit
or undermine any of an agency’s binding obligations under international or domestic law. Rather,
the AI Toolkit provides guidance, considerations and recommendations to be followed voluntarily.

Equally, the AI Toolkit is not intended to replace or subsume any of the management strategies,
policies or procedures already in place in law enforcement agencies, such as project management,
risk and needs analysis, data protection and assessment, or risk-benefit analysis. Instead, it is
designed to complement existing institutional and organizational processes and procedures: it is
most useful when used in conjunction with other relevant management tools.

7
 README File

Who is the AI Toolkit for?

The intended primary users of the AI Toolkit are personnel in law enforcement agencies, whether
local, regional or national. This includes a wide range of personnel – both officers and civilians
– in various departments and units, from the end users of AI systems in operational units to
innovation teams, legal divisions and procurement offices, as well as, in some cases, the chief of
police and their executive leadership.

However, it is also important to highlight the fact that ‘law enforcement’ can be a broad term.
For the purposes of the AI Toolkit, ‘law enforcement’ refers to the police and other similar state
authorities that exercise police functions such as investigating crimes, protecting people and
property, and maintaining public order and safety. In some cases, the term ‘law enforcement’
may include border control authorities, corrections officers, or counter-terrorism entities, for
example. In some cases, the judiciary, in particular prosecutors, is also considered to be part of
‘law enforcement’. While the AI Toolkit was not designed primarily with these latter kinds of ‘law
enforcement’ users in mind, it may nevertheless be of value for these authorities when exercising
police functions.

It is crucial to note that, apart from law enforcement personnel, many other non-law enforcement
stakeholders play a key role in implementing responsible AI innovation in law enforcement. These
include technology developers in the private sector or academia, civil society, the general public,
and other criminal justice actors such as the judiciary, prosecutors and lawyers – if they are not
already considered part of ‘law enforcement’. The AI Toolkit has been developed with the role of
these ‘secondary’ stakeholders in mind, and seeks to simultaneously increase general knowledge
and awareness of the underlying themes of responsible AI innovation in law enforcement,
demystify the issues involved, and facilitate more informed discussions among and between all
groups of stakeholders.

How to use the AI Toolkit?

The resources within the AI Toolkit have been specifically designed as standalone resources,
allowing each resource to be consulted independently of the others and at different times.

This approach takes into account the fact that each agency’s experience with AI, and their journey
towards responsible AI innovation, is unique. While some agencies may already be using AI

8
 README File

systems to support their day-to-day work and may have the corresponding in-house expertise,
other agencies may still be getting to grips with general knowledge and understanding of AI and
exploring the available technologies. This approach also allows for the fact that the AI Toolkit will
be relevant for many different types of law enforcement personnel at different stages of the life
cycle of an AI system.

RECOMMENDED PATHWAY THROUGH THE AI TOOLKIT

Although the resources within it are designed to be consulted independently, there is a

recommended pathway through the AI Toolkit.

9
 README File

This pathway begins with all AI Toolkit users – regardless of how familiar the agency
or individual is with AI – getting acquainted with the concepts upon which the AI

1
Toolkit is founded and the key terms that arise repeatedly throughout. With this in
mind, it is recommended that AI Toolkit users start by familiarizing themselves with
the Introduction to Responsible AI Innovation and the Principles for Responsible AI
Innovation, referring as needed to the Technical Reference Book for more technical
detail on AI.

Having established this foundation, it is recommended that each agency carries out a
self-assessment of its organizational readiness using the Organizational Readiness

2
Assessment Questionnaire, supported by the Organizational Roadmap. Again, it is
recommended that this assessment be conducted regardless of the level of experience
that an agency already has regarding the use of AI systems. This exercise will help
the agency to understand at what stage it is in terms of the necessary capabilities for
beginning or continuing the journey toward responsible AI innovation.

AI Toolkit users – particular those moving from a more general exploration of the
opportunities AI systems present toward the actual deployment of an AI system – are
then advised to move on to the Risk Assessment Questionnaire and the Responsible
AI Innovation in Action Workbook.

The Risk Assessment Questionnaire should be completed at the earliest opportunity

during the life cycle of an AI system. This will provide the teams in charge within
an agency with a solid understanding of the potential threats to individuals and

3
communities associated with a specific AI use case, thus allowing them to better align
their decisions with the Principles for Responsible AI Innovation. Given that assessing
risks is an iterative process, it is recommended that the Risk Assessment Questionnaire
be conducted periodically.

The Responsible AI Innovation in Action Workbook is designed to be an accompanying

resource for each of the various stages of the life cycle of an AI system – from planning
to deployment, use and monitoring, and anything else that may be involved – and for
the different teams within an agency that may play a role in these stages. AI Toolkit
users should regularly refer to the Workbook to help them put responsible AI innovation
into practice and to track and document their goals and progress.

The following table provides an overview of the main aims of each resource and a provisional
list of the users that may find the resources valuable. This is intended to provide further support
for agencies in determining when to consult each resource and who should be assigned the
responsibility of familiarizing themselves with its content.

10
 README File

RESOURCE PURPOSE USER

This resource provides Any individual or team in a

a basic overview of law enforcement agency
important AI concepts and that is interested in learning
Introduction to
an understanding of what more about the importance
Responsible
responsible AI innovation of a responsible approach to
AI Innovation
means, its foundations, and its AI innovation and wants to
significance in the context of understand the basis for such
law enforcement. an approach.

This resource serves as a

Any individual or team in a
foundation for the AI Toolkit.
law enforcement agency
The principles described
working with or in relation to AI
Principles for therein are aligned with
systems, including the chief of
Responsible policing principles to ensure
police, the executive leadership,
AI Innovation that AI systems are developed
and others such as decision-
and used for the benefit
makers in senior management
of society while protecting
positions.
human rights.

This resource seeks to

The chief of police and
help law enforcement
executive leadership, as well
agencies to understand
as decision-makers in senior
what elements should be in
management positions
place in order to pursue a
outside the executive
responsible approach to the
Organizational leadership, particularly those
development, procurement,
Roadmap in technology and innovation
and/or deployment of AI
units that are responsible
systems. Its goal is to support
for the deployment of AI
law enforcement agencies
systems.
with preparing to apply the
Principles for Responsible AI
Innovation.

11
 README File

The chief of police and

executive leadership, as well
This resource aims to support
as decision-makers in senior
law enforcement agencies to
Organizational management positions
determine how well equipped
Readiness outside the executive
they are to adopt responsible
Assessment leadership, particularly those
Questionnaire AI innovation, reflecting upon
in technology and innovation
the guidance provided in the
units that are responsible
Organizational Roadmap.
for the deployment of AI
systems.

This resource aims to help

law enforcement agencies
to identify and evaluate The teams or staff members
Risk
the risks that an AI system within an agency that are
Assessment
may pose to individuals and responsible for an AI system
Questionnaire
communities if the Principles at each stage of the life cycle.
for Responsible AI Innovation
are not sufficiently fulfilled.

This resource serves as

a basis to support law
enforcement agencies
in implementing and
The teams or staff members
Responsible AI operationalizing responsible
within an agency that are
Innovation in AI innovation and
responsible for an AI system
Action Workbook documenting the decisions
at each stage of the life cycle.
taken throughout the life cycle
of an AI system.

12
 README File

The resource is intended to

be used in conjunction with
the other AI Toolkit resources.
Technical It provides explanations of
Anyone using the AI Toolkit.
Reference Book key concepts and terms, as
well as associated fields, and
examples which are relevant
to law enforcement.

WHEN IS THE AI TOOLKIT APPLICABLE?

There are many different ways in which AI systems can be used in law enforcement, and some
are more likely to raise issues than others. In fact, with the growing prevalence of AI, it is highly
likely that many law enforcement agencies will already be making extensive use of AI systems or
systems that include some AI components. However, such systems will generally not have been
specially produced for or exclusively used in a law enforcement context. Some examples are the
use of email spam filters, navigation apps or even face ID to unlock service-issue smart phones.
The question that arises is therefore whether the guidance, recommendations and considerations
in the AI Toolkit apply to all law enforcement interactions with AI or only for specific systems,
tools and technologies that are above a particular ‘threshold’.

Given the rapid advancements in this field, the specific impact and risk of any single AI system
cannot be determined solely by looking at the nature of the use case itself or whether it was
especially developed for the law enforcement context. On the contrary, understanding and
managing the risks that may derive from implementing AI systems in law enforcement involves
a consideration of a diverse range of circumstances. In practical terms, this means that there
can be no definitive ‘threshold’ to establish when the AI Toolkit should be consulted or when its
guidance, recommendations and considerations should be acted upon.

Despite this, given their nature, the core principles for responsible AI innovation can be considered
applicable to all uses of AI systems and at all stages of the life cycle. They are general principles
designed to be adaptable to different contexts, and respecting them will contribute to promoting
good practices as agencies leverage the positive potential of using AI systems. With regard to the

13
 README File

applicability of the guidance, recommendations and considerations in the rest of the AI Toolkit,
agencies should consider the following three questions:

1. Does the use of the AI system pertain to the execution of the core functions of the law
enforcement agency, namely, to protect the community, prevent and investigate crime, and
ensure justice?
2. Having completed the Risk Assessment Questionnaire, have medium, high, or extremely
high risks pertaining to the AI system been identified?
3. In the event that the AI system has been identified as presenting a low risk, are there
any other potential effects on human rights that would necessitate considering the full
application of the AI Toolkit?

If the answer to any of these three questions is yes, then the recommendation is to take into
account the guidance, recommendations and considerations contained within the AI Toolkit to
the fullest extent possible.

How did the AI Toolkit come about?

The AI Toolkit was developed by INTERPOL and UNICRI in response to calls for support from
the global law enforcement community. The idea for the AI Toolkit initially emerged from the
first INTERPOL-UNICRI Global Meeting on AI for Law Enforcement in July 2018, during which
participating representatives from the law enforcement community flagged their concerns
regarding the challenges they faced in ensuring the responsible use of AI. Recognizing this
need, INTERPOL and UNICRI undertook the task of bridging the gap in the guidance for law
enforcement agencies on using AI responsibly and in a way which aligns with human rights and
ethical principles.

The AI Toolkit has been developed through a highly inclusive, consultative, and iterative process.
This process has been specifically designed with a view to ensuring that the AI Toolkit is technically
and operationally sound, promoting close interdisciplinary coordination with industry, academia,
criminal justice practitioners, civil society organizations and the general public, and fostering a
sense of transparency and a broader acceptance of the use of AI in law enforcement

The development process has included several notable components:

14
 README File

The Core Group of Experts:

INTERPOL and UNICRI established this group as an informal external review body to support
and guide the development of the AI Toolkit from a substantive perspective. It was comprised
of 56 experts from law enforcement, academia, the private sector, and civil society from over 30
different countries. The Core Group of Experts was convened in a virtual format on six occasions
during the development process.

Stakeholder Consultation & Engagement:

INTERPOL and UNICRI carried out regular consultations with representatives of the law
enforcement community throughout the development of the AI Toolkit. In addition, specific group
consultation sessions were organized with criminal justice practitioners, representatives from the
human rights community, and technology providers. When necessary, individual subject matter
experts/boards/organizations were also contacted on an ad-hoc basis for additional consultation
on specific topics. Finally, as public engagement, trust, and law enforcement transparency
play a vital role in the concept of responsible AI innovation, a global public opinion survey was
conducted to map public perceptions of the use of AI systems by law enforcement agencies,
and consultations were organized on specific public concerns around groups in vulnerable
circumstances. Feedback received during this process informed the drafting of the sections of
the AI Toolkit which address the importance of public engagement.

Peer Review:

Ahead of the AI Toolkit’s public release (V1.0), INTERPOL and UNICRI facilitated a technical review
of the AI Toolkit involving more than 30 experts with backgrounds in law enforcement, academia,
human rights, industry, criminal justice, the public sector, and international organizations from
across the globe. During this process, detailed feedback on each of the resources was received
and integrated into the AI Toolkit.

Testing the AI Toolkit in a Practical Setting:

To complement the technical peer review, the AI Toolkit has undergone testing by national and
local law enforcement agencies from 15 different countries, identified in response to a call for
volunteers issued by INTERPOL to all its member countries. Feedback received from testing
countries has been compiled and used to improve the AI Toolkit, resulting in the current revised
version.

15
 README File

ACKNOWLEDGMENTS
INTERPOL and UNICRI would like to express their deep appreciation to
the many experts from within the law enforcement community and other
national authorities, as well as industry, academia, civil society, and
international and regional organizations, that have devoted their time,
insights and perspectives to the process of developing the AI Toolkit.
Their contribution is invaluable and has been pivotal in shaping the AI
Toolkit and ensuring its content is technically accurate, comprehensive
from all perspectives, understandable, and implementable by law
enforcement agencies. INTERPOL and UNICRI would also like to
specifically thank the law enforcement agencies who participated in
testing the AI Toolkit, helping to ensure that it is a practical and useful
set of resources that will benefit the law enforcement community.
Without the collective contributions and collaboration of all those
involved, the AI Toolkit would not have been possible. This collaborative
effort is a testament to the strength of international cooperation, and it
demonstrates the shared commitment of all involved to ensuring that
responsible AI innovation is the future of policing.
How to cite this publication: UNICRI and INTERPOL. (Revised February 2024). Toolkit for
Responsible AI Innovation in Law Enforcement: README file.

© United Nations Interregional Crime and Justice Research Institute (UNICRI), 2024

© International Criminal Police Organization (INTERPOL), 2024

 README File

www.ai-lawenforcement.org

XVIII