Supporting responsible AI: discussion paper

21 August 2023

Australian Federal
Police submission
Supporting responsible AI: discussion paper

Department of Industry, Science and Resources

afp.gov.au
Australian Federal Police submission / 21 August 2023

Table of contents
Introduction 2

Threat Environment 3

Use of AI by the AFP 4

Management of the use of AI by the AFP 6
Governance 7

Internal policies and building oversight AI framework 8

Technical leadership 8

Training 9

Partnerships 9

Conclusion 10

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 1
Australian Federal Police submission / 21 August 2023

Introduction
1. The Australian Federal Police (AFP) welcomes the opportunity to make a submission to the
Department of Industry, Science and Resources’ public consultation on safe and responsible
Artificial Intelligence (AI) in Australia. This submission addresses both identified criminal
threats relating to AI, and considerations for the use of AI for law enforcement purposes.

2. The AFP protects Australians and Australia’s interests from a wide range of harms by virtue
of our diverse functions. As the Commonwealth’s policing agency, the AFP has responsibility
for enforcing Commonwealth criminal law, contributing to combating complex, transnational,
serious and organised crime impacting Australia's national security and protecting
Commonwealth interests from criminal activity in Australia and overseas. The AFP also has
responsibility for providing policing services to the Australian Capital Territory (ACT) and
Australian external territories.

3. Technology is an enabler for the broader community including criminal enterprises, and in
turn for policing. The AFP’s Technology Strategy recognises that all crime types and policing
functions are affected by the use of technology. The emergence of increasingly mature AI
and machine learning (ML) has accelerated this impact as it permeates many facets of
technology and society.

4. The AFP recently published the “Blue Paper: To 2030 and Beyond” detailing the future of
federal policing. This sets out our strategic approach to safeguarding lives, livelihoods, and
Australia’s way of life well into the 2030s. Within this strategic outlook, the Blue Paper
acknowledges the significance of Digital Evolution 4.0, including the influential role of AI, as a
major driver of transformative changes affecting the Australian community and in turn
policing. The AFP is committed to five future core priorities, which includes leadership in
embracing and leveraging technology. To meet the ever evolving operational and threat
environment, successful implementation of new and emerging technology will play an
important role in shaping the AFP’s future operational effectiveness and efficiency.

5. The AFP recognises technology is also a key enabler for criminal activity against Australians,
including by cyber criminals, the online grooming and subsequent abuse of children, and
easily accessible exposure to abhorrent and violent extremist material, radicalising our most
vulnerable - including children. The effect is delivered from a distance, and at speed and
scale that is unprecedented in our history. AI and ML delivers further opportunities for
criminals to scale their activities, yet these tools are equally essential to law enforcement
responding to this threat.

6. The AFP is determined to understand and engage with the threats and opportunities posed
by AI. Into the future, the use of AI will be required for the AFP to effectively achieve our
mission, to uphold public safety and combat criminal activities. To do this, the AFP will
assess the following opportunities:
 continually explore and seek opportunities to invest in new technology;

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 2
Australian Federal Police submission / 21 August 2023

 invest in / upskill our people to ensure they have the requisite skills and
understanding of AI;
 ensure appropriate processes and governance in relation to the implementation of
new technologies to meet community expectations;
 continue to strengthen, and foster new, relationships across government, industry
and academia; and
 work with the broader community to ensure Australia has a fit for purpose regulatory
environment.

Threat Environment
7. Although AI offers great value to humanity, criminal enterprises have been, and will continue
to be early adopters of technology wherever they see an opportunity to advance their
insatiable appetite to benefit at the expense of others. Criminals will use any means possible,
even if it means undermining legal principles or basic democratic rights, or creating a feeling
of terror within a community. The policy environment needs to be as agile and dynamic as
those seeking to cause harm in our communities.

8. The AFP considers the key threats posed by AI affecting the criminal environment include:
 Increased Potency: AI enables more frequent and widespread attacks, amplifying their
impact.
 Enhanced Accessibility: AI lowers the entry bar and cost for non-technical individuals
to engage in malicious activities.
 Exploitation of Human-Centric vulnerabilities: AI is more efficient and effective in
leveraging vulnerabilities unique to human behaviour.
 Deliberate sabotage of critical algorithms: AI introduces the risk of poisoned and
sabotaged algorithms leading to ineffectual use of AI and/or harm.

9. The threats of future generations of AI tools may create unforeseen consequences that
further enable criminal activity. By evaluating the current capabilities of AI, we gain some
insight into future risks that may be realised as the criminal enterprise takes further
advantage of the AI capability as it develops. These technology-facilitated crimes could
become increasingly prevalent and have a profound impact on our communities as new
threats emerge.

10. One example of threat faced now is deepfakes, which involve manipulating audio and video
to impersonate a person. Deepfakes can and have been used to discredit public figures,
extort funds, and influence democratic processes. This has serious implications for trust,
reputation and public safety. The sophistication of deepfakes makes it difficult to distinguish
‘fact from fiction’ and the ‘real from virtual’, and poses a real challenge for both law
enforcement and the broader community. In turn, this could lead to widespread distrust of
audio and visual content, which is an important way that governments and other bodies
connect and communicate with their communities.

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 3
Australian Federal Police submission / 21 August 2023

11. Another recognised form of harm is the creation and distribution of child exploitation
material. The creation of AI generated (fake) child exploitation material is not only illegal but
also encourages and normalises the abuse of real children and diverts law enforcement
resources invested in victim identification.

12. The rise of AI-driven cybercrime is another key identified harm to the community. Criminals
or foreign entities can leverage AI algorithms to launch large-scale cyber-attacks that are
more sophisticated and targeted, such as phishing scams, ransomware and data breaches.
Over $33 billion was reported as lost from cybercrime in Australia in 2020-21 and this is only
expected to increase. The ability of AI to develop and distribute convincing phishing emails
that are indistinguishable from genuine emails or that are designed to circumvent automated
detection systems presents a new frontier to identity theft, financial fraud and how online
scams are orchestrated.

13. Just as AI technology evolves over time, so will the threat. The challenge of countering new
threats will require continual investment in technology by law enforcement, supported by
training and governance, to enable a technically literate and empowered workforce. The
fundamental focus of law enforcement’s use of AI must be the effective balancing of
transparency, accountability, fairness, privacy and security, proportionality and justifiability.

14. Changes to legislation and regulations need more detailed consideration to ensure the
Australian community is protected. The AFP welcomes engagement to ensure policy
changes are practical, proportionate and enforceable. Striking the right balance between the
role of police in safeguarding our community, and the safeguarding of individual rights
requires robust consideration, governed by general and sector-specific regulations. Adhering
to the Peelian principles of policing, the AFP will prioritise community engagement and
transparency in AI development, acknowledging that successful policing within democratic
societies is based on trust to apply powers fairly and without fear or favour.

15. Protecting our community from harm through strong relationships and enduring
partnerships with key stakeholders is at the heart of achieving operational outcomes. The
AFP is increasingly confronted by evolving, multi-disciplinary challenges that require multi-
agency and multi-jurisdictional solutions. The same principle applies to combatting the
threats emerging from the use of AI. The AFP will leverage our domestic and international
networks to further strengthen partnerships across the country and the world.

Use of AI by the AFP
16. To date the AFP has taken a cautious approach to harnessing the potential of AI despite its
early adoption and use by those with intent on harming Australians. We recognise the need
to engage more on the potential of AI to remain effective in our mission to uphold public
safety and combat criminal activities. The AFP acknowledges that we must hold ourselves to
a higher standard than those of our adversaries and be cautious in the use of broader private
industry offerings due to the ethical and privacy impacts of the technology.

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 4
Australian Federal Police submission / 21 August 2023

17. The AFP is continuing to evolve our internal processes to navigate the development and
deployment of emerging technologies. As the technology landscape changes, we will need to
ensure accountability, transparency, and responsibility when adopting any new technology.
This principle has been, and will continue to be, the cornerstone of the AFP’s journey towards
responsible and ethical innovation.

18. Managed correctly, investments in AI will propel our organisational capabilities, enabling us
to address a changing and increasing threat environment intent on harming Australians and
Australia’s way of life. The scale of technology-facilitated crime continues to grow with AI
enabling criminality on a truly industrialised scale. Ransomware attacks on Australian
businesses will increase in both frequency and severity, with an annual increase of 15% on
reported attacks from the previous year. Within this environment, law enforcement resources
will continue to be challenged by criminal enterprises we are attempting to contain. To
counter this, successful adoption of AI will be critical. AI offers the AFP opportunities to:

a. Create operational efficiencies in information discovery and understanding;
b. Improve situational awareness to inform better human decision making; and
c. Minimise physical and psychological risks to AFP capabilities and members.

19. The AFP’s current utilisation of AI has generally been limited to facilitating the transformation
of data from one format to another, to enhance analysis and processing needs. AI tools such
as Large Language Models (LLM) and broader Neural Networks present the AFP with an
opportunity to simplify the task of identifying potential value from large lawfully collected
datasets. By speeding the discovery task, members can make decisions earlier and execute
the necessary actions accordingly.

20. An example of AI’s practical application in law enforcement investigations is the use of AI-
based translation technology, which enables the translation of foreign materials into English.
By combining the strengths of AI and human expertise, we achieve a streamlined and
effective process, significantly enhancing our ability to handle large volumes of multilingual
data.

21. AI adoption is swiftly transforming both communities and private industries with whom we
partner to combat crime, driving innovation and efficiency across diverse sectors. The
banking sector, for example, has rapidly embraced AI technology to enhance both customer
experience and security measures. AI’s analytical capability is harnessed to detect and
prevent fraudulent activities in real-time. By analysing vast amounts of transactional data, AI
can swiftly identify irregular patterns such as money laundering and flag potential fraud
ensuring customer safety and financial integrity. The AFP will work with industry partners
and community organisations to identify opportunities to learn, while concurrently fostering
educational initiatives about AI’s implications and associated risks. This knowledge will help
our community to protect themselves from crime, enhancing overall crime prevention
resilience.

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 5
Australian Federal Police submission / 21 August 2023

22. In parallel, the AFP has been working in close collaboration with the AI for Law Enforcement
and Community Safety Lab (AiLECS Lab) at Monash University, to not only develop AI
technologies but also to develop the underlying frameworks to ensure we apply AI in an
ethically transparent and accountable manner. The use of AI technologies to assist in the
identification and classification of child exploitation material is one example. The primary
objectives are to enhance the ability of all law enforcement agencies and non-government
organisations to effectively and efficiently locate victims while minimising the detrimental
impact of reviewing explicit material. By employing AI technologies, the reduction of harm
extends beyond victims, to those tasked with identifying and locating them, ultimately
yielding more favourable outcomes for all stakeholders.

23. A key pillar of our criminal justice system is that the burden of proof remains with the
prosecution. The prosecution must be able to satisfy the court of the authenticity, reliability
and accuracy of evidence. Any process for obtaining and handling evidence must not
undermine these principles. This remains an ongoing key point of collaboration between the
AFP, the Commonwealth Director of Public Prosecutions, AiLECS Lab, and our law
enforcement partners.

24. By forging strong and lasting collaborations with our partners and community stakeholders,
we can harness the potential of AI in policing. These enduring partnerships foster an
exchange of knowledge, cutting-edge innovations, standards, and build trust, enabling us to
continually improve information discovery, accelerate operational speed, and reduce risks to
the community and AFP members.

Management of the use of AI by the AFP
25. At the heart of the AI conversation is the race between the rapid development and adaptation
of technology and the need for prudent governance and accountability, informed by
community expectations. As a law enforcement agency charged with protecting Australians
and Australia’s interests, the AFP cannot afford to stand still while adversaries adopt this
technology to obtain a competitive advantage and cause unprecedented harm. The AFP
understands any innovation we deploy must withstand scrutiny and align with societal values
and public expectations. This is at the core of our need to maintain transparency.

26. As a responsible and trusted policing organisation, the AFP will lead by example developing
AI technologies through collaboration and engagement with all stakeholders to
comprehensively understand future implications. The key is to proactively undertake due
diligence into technologies before deployment, necessitating thoughtful planning, consistent
ethical considerations, robust governance and oversight to ensure that AI serves the greater
good and aligns with societal values.

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 6
Australian Federal Police submission / 21 August 2023

27. The AFP supports ongoing engagement with the community to ensure Australia has the right
governance settings to respond to the rapid development of AI. This extends to the
complexities of harmonising governance frameworks with those used globally or by our
foreign law enforcement agency (FLEA) partners, to enable a shared commitment to ethical
AI practices, ultimately enhancing law enforcement capabilities across borders. The AFP will
continue to leverage law enforcement structures such as the, Five Eyes Law Enforcement
Group (FELEG) Interpol and EUROPOL to navigate the complexities of sector-specific AI
utilisation. Specifically, our organisation remains dedicated to implementing the Australia
New Zealand Policing Advisory Agency (ANZPAA) AI Principles that have agreed by all
Australian and New Zealand Police Commissioners.

28. The dual-use nature of AI presents significant challenges for policing, raising concerns about
potential misuse and its impact on public safety and trust. AI technologies offer both
beneficial and harmful applications, making regulation complex. The use of AI for policing
raises ethical and legal questions on privacy, bias and accountability. The seamless
integration of AI, imperceptible to the human eye, fundamentally alters the security
landscape by enabling the capability to comprehend information beyond an individual
human’s capability.

29. The AFP recognises AI is another tool, and it will not replace the requirement a human must
remain accountable for any decision that affects on the rights of another human. While AI
can assist decision-making, the AFP recognises policing is a human and societal contract
that will always require human judgement and interaction. Policing is deeply connected to
society and must reflect the values, norms and expectations of the community it serves and
critically requires human oversight and accountability. While industry may be moving
towards granting autonomy to AI for operational efficiency, the AFP will use AI as an enabler,
not a decision-maker. Maintaining human involvement in decision-making is a crucial
differentiator for the police, ensuring responsible and ethical AI adoption to prioritise the
wellbeing and safety of the community and our members.

30. To achieve this, the AFP is prioritising the development of five key enabling aspects of AI
technologies:
a. Governance (accountability and human oversight)
b. Internal policies and frameworks (responsibility)
c. Technical leadership (technical excellence)
d. Training (competency in use and trust in technologies)
e. Partnerships (collaboration and transparency)

Governance
31. The AFP is prioritising effective governance to guide the implementation of AI technologies,
adhering to general regulations and legal frameworks that apply across industries. In
particular, the AFP will ensure compliance with the Privacy Act. By reporting on AI initiatives,

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 7
Australian Federal Police submission / 21 August 2023

data usage, and privacy measures, the AFP will demonstrate its commitment to responsible
AI practices reflective of community expectations and values.

32. The AFP will continue to remain accountable through reporting to oversight agencies,
including the Commonwealth Ombudsman, to ensure our use of technology is aligned and
adheres to relevant legislation and sector-specific regulations.

Internal policies and building oversight AI framework
33. Through enhanced compliance and governance frameworks, the AFP will ensure ethical and
responsible AI use into the future. These enhancements in line with the Commonwealth
Government’s legislative and policy requirements and human oversights will continually be
strengthened through our agency’s commitment to continuous improvement.

34. To strengthen responsible AI use, the AFP will prioritise the establishment of an AI Oversight
Framework aligned with standards, facilitating repeatability, validation, verification, and both
internal and external assessment. This will be built into the AFP’s existing strategic
governance arrangements. As we have demonstrated in building world-leading Forensics
capabilities and frameworks, the AFP will apply the same approach by embodying
robustness and oversight, fostering trust at organisational, jurisdictional, and community
levels. By adhering to this comprehensive framework, the AFP will ensure the integrity of AI
applications and other emerging technologies (such as those in the space and robotics field)
instilling confidence in our stakeholders and reinforcing our commitment to responsible
adoption of emerging technology.

35. Recognising the dynamic nature of risk management and post-deployment oversight for AI, it
is crucial the AFP’s risk management approach is continuously iterative and runs throughout
the entire lifecycle of any AI system. The significance of regular, systematic updates and
comprehensive audits cannot be overstated. As the technology evolves and the environment
in which AI operates shifts, staying attuned to potential risks and ensuring compliance with
ethical standards demand constant vigilance.

Technical leadership
36. Technical leadership plays a vital role in AFP’s responsible adoption of AI. Our technical
leaders and specialists possess the expertise and skills to guide the implementation of AI
technologies. Technical leadership ensures AI is governed, developed and deployed ethically,
addressing biases and risks, while driving innovation and maintaining compliance with
regulations.

37. The AFP has conducted strategic workforce planning, which sets out our key areas of focus
for the next five years, committing to actions through a series of targeted strategies. The
AFP will commit to delivering a workforce that is skilled for today and tomorrow, configured
for operational agility, comprised of engaged and supported employees and shaped by
contemporary, data-driven strategic decisions. The implementation of this strategy will play a

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 8
Australian Federal Police submission / 21 August 2023

pivotal role in shaping the AFP’s technical leadership, ensuring the AFP has the right level to
lead and navigate the complex landscape of AI.

Training
38. The AFP will explore utilising internal and external training programs to equip our entire
workforce, including frontline members, with the essential skills to utilise AI technologies
appropriately. It is essential all employees can recognise where AI exists and understand its
limitations and risks, and determine its appropriate use in accordance with internal and
externally adopted policies and frameworks, in combination with their own knowledge and
skills. Having comprehensive training programs will empower all our employees to
responsibly and confidently use AI technologies, cultivating a culture of accountability and
trust in the benefits of AI for policing.

39. As AI technologies are integrated into policing practices, our frontline members may
encounter situations where AI-driven evidence becomes crucial in court proceedings.
Equipping the workforce, including frontline members and specialists with essential AI skills
will become paramount, making complex AI mechanisms understandable and transparent to
the court and the parties involved.

Partnerships
40. In our policing efforts, partnering with industry and comprehending the AI technologies we
acquire is vital. This collaboration ensures the AI tools align with our ethical standards and
policing objectives. Understanding the inner working of AI systems empowers us to make
informed decisions, considering potential biases and limitations.

41. The AFP will continue to strengthen our partnerships with academia and the AI industry
through transparent dialogue and foster collaborative relationships, aiming to harness AI for
policing purposes effectively. Our partnership with Monash Universities AiLECS Lab
exemplifies this, with its objective of creating a safer community through effective, ethical
and transparent AI use.

42. The AFP notes international partnerships present significant opportunities for AI adoption
through facilitating the exchange of knowledge and capabilities. Concurrently, it is vital that
the AFP assess governance frameworks with those used globally or by our key FLEAs.
Achieving alignment while respecting differing legal and societal frameworks is crucial for
responsible AI adoption. Collaboration and transparency with our international partners
foster trust and enable a shared commitment to ethical AI practices, ultimately enhancing
law enforcement capabilities across borders.

43. The AFP will leverage existing structures such as the ANZPAA, FELEG, Interpol and
EUROPOL to assess governance frameworks and address global AI complexities and sector-
specific utilisation.

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 9
Australian Federal Police submission / 21 August 2023

Conclusion
44. Investments in AI will propel our organisational capabilities, enabling us to address a
changing threat environment harming Australians and Australia’s way of life. Managed
correctly, the AI will offer the AFP opportunities to create operational efficiencies, improve
situational awareness to inform better human decision making, and minimise risks to the
public safety, AFP members and capabilities.

45. Responsible adoption of AI in policing requires a multifaceted approach, which the AFP is
assessing the opportunities to strengthen five key enabling components: governance,
policies, technical leadership, training and partnerships.

46. Collaborating internationally demands transparent dialogue to assess governance
frameworks and address diverse legal and ethical considerations. Remaining accountable is
vital, and achieved through oversight governance and regulations that uphold and demand
responsible AI practices. Building an AI Oversight Framework aligned with international
standards and supported by internal policies will provide clarity and guidance in ethical AI
adoption.

47. Fostering technical leadership in the AFP workforce empowers informed decision-making,
innovation, and effective implementation, ultimately enhancing policing capabilities and
ensuring public safety in our communities.

48. The AFP will explore opportunities to equip our workforce with comprehensive training to
ensure adept navigation of AI technologies, instilling trust in their use and fostering a culture
of accountability. Partnering with industry will enable the AFP to understand AI technologies,
ensuring alignment with our ethical standards and policing objectives – with the
community’s expectations of us at the front of mind.

49. The AFP welcomes the opportunity to engage further on regulatory and governance
responses to create safe and responsible AI in Australia.

AFP / MANAGER TECHNOLOGY STRATEGY & DATA | CHIEF INFORMATION OFFICER 10