Supporting responsible AI: discussion paper

Department of Industry, Science and Resources, Technology Strategy Branch
By email: DigitalEconomy@industry.gov.au

Wednesday, 16 August 2023

Dear Minister Husic,

The Digital Industry Group Inc. (DIGI) thanks you for the opportunity to provide our views on the Safe and responsible AI in Australia Discussion paper.

By way of background, DIGI is a non-profit industry association that advocates for the interests of the digital industry in Australia. DIGI’s founding members are Apple, eBay, Google, Linktree, Meta, Snap,
Spotify, TikTok, X (f.k.a Twitter), and Yahoo. DIGI’s vision is a thriving Australian digitally-enabled economy that fosters innovation, a growing selection of digital products and services, and where online safety and privacy are protected.

DIGI welcomes the work of the Department of Industry, Science and Resources to realise Australia’s potential as a responsible and inclusive Artificial Intelligence (AI) leader, as the development of digital technology is essential to Australia’s continued economic prosperity. The Australian Productivity
Commission’s 5-year Productivity Inquiry: Advancing Prosperity report found that ‘harnessing data, digital technology and diffusion and creating a more dynamic and competitive economy’ were all key themes required to ensure Australia’s economic growth.1

There are significant economic and social opportunities in the development of new technologies such as
Artificial Intelligence (AI) and its applications, such as Automated Decision Making (ADM), large language models (LLM), and Multimodal Foundation Model (MfM). Domestic regulatory settings will directly influence the ability for Australia to access these opportunities and remain competitive on the global stage. In 2019, Australia’s Digital Opportunity, a report produced by AlphaBeta (now Accenture) and commissioned by DIGI.2 It quantified the extraordinary contribution of Australia’s technology sector to the national economy. It found that, at that point in time, the technology sector contributed $122 billion each year to the national economy, or 6.6% of GDP. A subsequent estimate by Accenture in 2021 found that the tech sector contributed $167bn, or 8.5%, of GDP, demonstrating the rapid growth of the sector.3 However,
Australia has not fully leveraged the economic opportunity of the technology sector and, in 2019, ranked second last in the OECD for gross value added for the size of its technology sector. Looking ahead, it is essential that the Government strikes the right balance in regulatory response to create the conditions where Australia can remain competitive on the global stage and take full advantage of the opportunities presented by AI.

DIGI supports development of AI, and its applications, underpinned by strong safety principles and a conscious assessment of social and economic benefits. We note that this work is already well underway by industry and can be seen in many existing internal governance measures to deliver trustworthy and safe AI technology, such as internal and external security testing before the release of new AI systems or

1
Productivity Commission (2023), 5-year Productivity Inquiry: Advancing Prosperity, Available at: https://www.pc.gov.au/inquiries/completed/productivity/report
2
Unless otherwise noted, all statistics from this section are from AlphaBeta (2019), Australia’s Digital Opportunity, accessed at: https://digi.org.au/wp-content/uploads/2019/09/Australias-Digital-Opportunity.pdf
3
Accenture (2021), The economic contribution of Australia's tech sector, accessed at https://techcouncil.com.au/wp-content/uploads/2021/08/TCA-Tech-sectors-economic-contribution-full-res.pdf
voluntary research to understand and prevent system bias. AI can support better decision-making, public safety and more inclusive and informed societies. DIGI also acknowledges the capacity for unintended consequences in perpetuating biases and other risks. We believe that a coordinated, proportionate response across government can address potential harms and mitigate risk related to AI, while supporting local innovation.

Our main arguments in this submission are:
1. A proportionate, balanced approach to AI regulation is key to mitigating and addressing
potential risks related to the use of AI, while accessing the technology’s significant
socio-economic benefits.
a. DIGI agrees with adopting a proportionate, risk-based framework to prevent and address
issues related to the use of AI. DIGI recommends that the Department’s
recommendations should address the mitigation of potential risks by identifying the
defined harms or challenges and balancing responsibility between the upstream
designers of AI models and technology with the downstream deployers.

2. Regulatory analysis to produce sector-specific guidance and gap analysis to inform
legislative reform process is needed, as well as voluntary processes to better account for
potential risks related to the application of AI.
a. DIGI recommends that policy responses first build on existing regulation, rather than
introducing new legislation aimed at regulating AI as a technology. Assessing the use of
AI within sectorally relevant regulation will allow for the most contextually relevant
understanding of the application of the technology and its potential risks.
b. DIGI recommends the Government, including specific regulators, be resourced to assess
to what extent AI is already regulated within the scope of existing laws in all the sectors
where AI will be applied. For example, this should include relevant laws in healthcare,
insurance and – in relation to the digital industry – the application of the Online Safety
Act 2021 (OSA) to address AI proliferated online harms.
i. The sector-specific guidance should be well-socialised by Government through a
program of measurable outreach in relevant industries in order to ensure uptake.
c. This assessment should inform sector-specific guidance that those regulators issue to
the industries that they regulate around the application of existing laws to AI technology,
noting that the context in which AI is employed is key to understanding risk. This
sector-specific guidance might encourage company policies or voluntary principles to
address any gaps identified where existing regulation does not address identified harms
proliferated through AI.
d. Gaps identified through this analysis should feed into existing and future reform
processes (such as the Privacy Act Review) to identify what can be applied to this
emerging use of technology.

3. A focus on Australia's economic competitive advantage, international interoperability and
coordination.
a. Regulatory settings and reform proposals across various regulators and Departments
have direct implications on Australia’s competitive advantage and economic prosperity.
i. For example, Australia’s current based fair use exceptions for copyright
infringement are potential barriers to the training and development of AI models
(which often require the input of copyrighted materials), which could ultimately
stifle local innovation.
b. There are significant social and economic benefits in the adoption and development of AI
that the Government should be cautious of limiting the ability of Australia to take full
advantage of these opportunities as an unintended consequence of a well meaning rush
to regulation.

1
i. The Government should also consider the risks present in not fostering local
innovation or development of AI technologies and the broad economic impacts
including to Australia’s long term prosperity and productivity.
c. We recommend that the Department ensures consistency and coordination in AI
governance through a whole of Government response and clear guidance for regulators
to reduce undue burden on industry and prevent barriers to innovation.
d. Global governance of AI is in flux and DIGI recommends that the Government remains
cautious of replicating any one international model until its efficacy can be better
assessed.
e. AI technologies are being developed at a global scale. Any regulatory developments
should carefully consider all economic implications, including the competitiveness of
Australian business, the ease of conducting international business, and implications with
Australia’s international trade partners. A heavy focus on domestic regulation without a
clear consideration of international cooperation and harmonisation across multiple
jurisdictions could create conditions where economic opportunities are missed without a
meaningful reduction in risk.

We thank you for your consideration of the matters raised in this submission, and we look forward to further discussion with you. Should you have any questions, please do not hesitate to contact me or my colleague Tahlia Davies.

Best regards,

Sunita Bose
Managing Director, DIGI sunita@digi.org.au

Table of contents
A proportionate, balanced approach to AI regulation is key to mitigating and addressing harms related to the use of AI, while accessing the technology’s significant socio-economic benefits. 3
Definitions 3
Understanding the opportunities for Australia in AI technologies 3
The role of AI in online safety 5
Regulatory analysis to produce sector-specific guidance and gap analysis to inform legislative reform process is needed, as well as voluntary processes to better account for harms related to the application of AI. 6
Analysis of existing laws and producing sector specific guidance 6

2
A focus on Australia's economic competitive advantage, international interoperability and coordination. 8
Clarity, coordination and cohesion across Australia’s AI governance 8
Australia’s governance response and international models 8
Conclusion 10

A proportionate, balanced approach to AI regulation is key to mitigating and addressing harms related to the use of AI, while accessing the technology’s significant socio-economic benefits.

Definitions
As the Discussion Paper notes, there is no single agreed definition of AI. DIGI notes that the fast-paced evolution of the technology presents some challenges in reaching a detailed definition, as it can quickly become outdated in line with new developments. For clarity and coordination across domestic and international regulatory schemes, DIGI suggests that the Government adopt a common lexicon based on the OECD AI principles, upon which Australia engaged in the development.4 For example, the principles outline the definition of an AI system as a ‘machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments. AI systems are designed to operate with varying levels of autonomy’.5 While the difference between this definition and that included in the Discussion Paper is subtle, adopting a common lexicon with other OECD countries could prove useful for future international engagement and consultation with
Australia’s allies and partners.

Understanding the opportunities for Australia in AI technologies
The World Intellectual Property Organisation notes AI as an example of a ‘frontier technology’.6 These are rapidly emerging technologies that have enormous potential for both driving economic growth and addressing pressing social challenges.

How to access these opportunities while mitigating against risks and potential misuse is a question that governments, industry and civil society are grappling with around the globe. In its recent 2023 Global
Education Monitoring Report, UNESCO summarised the global nature of this pursuit for balance in AI governance as ‘a consensus (is) forming about the need to enjoy AI’s benefits while eliminating risks from its unchecked use, through regulation relating to ethics, responsibility and safety’.7 This submission will further explore DIGI’s position on finding this balance.

Firstly, DIGI believes AI can support better decision-making, public safety and more inclusive and informed societies, in part because its deployment through algorithms and machine learning is already being used by a wide diversity of private sector industries and public sector departments for social benefits. Examples of the myriad of beneficial applications include:

4
Ibid.
5
OECD (2029), Ahttps://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0449
6
World intellectual Property Organization ‘ What are frontier technologies; fact sheet’ 2023
7
UNESCO (2023) Global education monitoring report summary, 2023: technology in education: a tool on whose terms?
Available at: https://unesdoc.unesco.org/ark:/48223/pf0000386147

3
○ Benefits to health: AI is helping people attain better health and well-being; a report by
PwC demonstrates how AI is already transforming eight components of the healthcare
system, including preventative health, diagnosis, decision-making, palliative care,
research and training8. As one example, Google’s hearing technology partnerships are
harnessing AI to develop listening and communications technology.9

○ Disability access and services: AI is transforming inclusion and access to services for
people with disabilities and the elderly. AI-powered devices that use voice commands,
such as Google Home and Google Assistant technology10 are being used by people with
limited sight or mobility11, and Meta uses AI to automatically write photo captions for the
blind and visually impaired.12

○ Environmental analysis and crisis planning: AI is being used to better predict impacts of
environmental changes or natural disasters. For example, AI is being deployed to and
support climate resilience planning.13 Google is also supporting research and
development of flood forecasting through its Impact Challenge programme.14

○ Entertainment and cultural material: Machine learning has long powered curation and
recommendations for videos, music, and textual works. For example, AI is used by music
streaming services to organise hundreds of millions of sound recordings and offer users
custom playlists, trivia, and other types of personalised music listening sessions.

DIGI also considers that there are significant economic opportunities in the development and adoption of
AI. In its recent Australia’s Generative AI opportunity report, the Tech Council of Australia estimated that generative AI alone could contribute $45B annually to the Australian economy by 2030. In the medium and fast paced scenarios, this figure could be $75B or $115B respectively.15 This highlights both the potential benefits of fostering local innovation and AI adoption and the opportunity cost of not doing so.
These are illustrative examples to demonstrate that there are significant socio-cultural opportunities in the development of AI technologies, as well as the potential to further grow Australia’s economic prosperity.

The role of AI in online safety
Beyond the potential social and economic benefits, DIGI notes that AI also plays a key role in delivering relevant services and online safety mechanisms for many digital platforms. On relevant large digital platforms, AI and algorithms play an important role as a sorting mechanism for the millions of terabytes of information online, enabling people to readily obtain relevant content and information.

8
PwC (June 2017) What doctor? Why AI and robotics will define New Health, available at: https://www.pwc.com/gx/en/industries/healthcare/publications/ai-robotics-new-health/transforming-healthcare.html.
9
Google, Five New Partnerships (March 2023), available at: https://blog.google/intl/en-au/company-news/technology/ai-hearing-initiative/
10
Feros Care (2019), “MyFeros and Google Assistant are helping seniors live in their homes longer” available at https://www.feroscare.com.au/feros-stories/articles/myferos-and-google-assistant-are-helping-seniors-live-in-their-h omes-longer
11
The Tipping Foundation (2018) 6 ways smart home technology is benefiting people with disability, available at: https://www.tipping.org.au/6-ways-smart-home-technology-is-benefitting-people-with-disability/
12
Matt Burgess (April 5, 2016) “Facebook's AI now writes photo captions for blind users”, Wired UK. Available at https://www.wired.co.uk/article/facebook-ai-image-recognition-caption-accessibility-blind-users
13
Nature (January 8, 2023) AI for climate impacts: applications in flood risk, Available at: https://www.nature.com/articles/s41612-023-00388-1
14
Google Impact Challenge, Available at: https://impactchallenge.withgoogle.com/globalgoals/projects/flood-forecasting
15
Tech Council (2023), Australia’s Generative AI opportunity, Available at: https://techcouncil.com.au/wp-content/uploads/2023/07/230714-Australias-Gen-AI-Opportunity-Final-report-vF4.pdf

4
AI is also used to safeguard the safety and security of Internet users, and to address harmful content.
Such technology is having a positive effect; In Q1 of 2023, approximately 99.4% of the comments removed from YouTube were detected by automatic flagging. Automated flagging also allows videos that violate community guidelines to be removed before they are widely viewed. In the same quarter, 42.8% of the videos flagged by automation were removed before they received a single view.16 Similarly, Meta uses artificial intelligence to proactively detect harmful content before it is seen by users. For example, Meta proactively detected 98.3% of child exploitation material in Q1 2023 before a user reported it.17Proactive detection is also being used extensively to identify and prevent scams targeting Australian consumers.
For example, Gmail proactively blocks more than 99.9% of spam, phishing, and malware before it reaches users.18

The use of algorithms to promote online safety is consistent with the Government’s expectations of industry. For example, the draft BOSE determination, that came into effect with the Online Safety Act
(OSA) on January 23, 2022, identifies the detection of material and activity as a reasonable step service providers can take to ensure end users are safe.19 The Office of the eSafety Commissioner’s Safety by
Design principles include “Using scanning and filtering technology to ensure user safety is upheld on the site and users are not exposed to inappropriate or sensitive content.”20The registered Social Media
Services Online Safety Code (Class 1A and Class 1B Material), developed under the OSA, also includes use of artificial intelligence and machine learning as safety measures related to the detection and removal of child exploitation and pro-terror material.21

It is important to note that algorithms also do not operate in isolation from human intervention; in relation to content removal, it is often the case that AI surfaces problematic content for a human moderator to review for context and accuracy, and to guide the most effective decision. AI plays an important role in scanning content at a scale that humans could never achieve, at a speed which was previously not possible. It forms a key part of how online safety challenges are addressed at a large scale.

Regulatory analysis to produce sector-specific guidance and gap analysis to inform legislative reform process is needed, as well as voluntary processes to better account for harms related to the application of AI.

Analysis of existing laws and producing sector specific guidance
The Discussion Paper notes that AI can already be addressed by a number of existing laws. DIGI recommends that policy responses first build on existing regulation, rather than introducing new legislation aimed at regulating AI as a technology. Assessing the use of AI within sectorally relevant

16
YouTube (2023), YouTube Community Guidelines enforcement, accessed at https://transparencyreport.google.com/youtube-policy/removals
17
Meta (2023), “Community Standards Enforcement Report Q1 2023”, accessed at https://transparency.fb.com/data/community-standards-enforcement/child-nudity-and-sexual-exploitation/facebook/
18
https://workspace.google.com/blog/identity-and-security/how-gmail-helps-users-avoid-email-scams
19
Department of Infrastructure, Transport, Regional Development and Communications, Draft Online Safety (Basic
Online Safety Expectations) Determination 2021, accessed at https://www.infrastructure.gov.au/have-your-say/draft-online-safety-basic-online-safety-expectations-determination-2
021-consultation
20
Office of the eSafety Commissioner, “Safety by Design | Principles and background”, accessed at https://www.esafety.gov.au/industry/safety-by-design/principles-and-background
21
Schedule 1 – Social Media Services Online Safety Code (Class 1A and Class 1B Material (2023) https://onlinesafety.org.au/wp-content/uploads/2023/06/230616_1_SMS-Schedule_REGISTERED-160623.pdf

5
regulation will allow for the most contextually appropriate understanding of the application of the technology and its potential risks, while creating settings that foster innovation and the socio-economic benefits of the technology.

DIGI recommends the Government, including specific regulators, be resourced to assess to what extent AI is already regulated within the scope of existing laws in all the sectors where AI will be applied. For example, this should include relevant laws in healthcare, insurance and – in relation to the digital industry
– the application of the Online Safety Act 2021 (OSA) to address AI proliferated online harms. The sector-specific guidance should be well-socialised by Government through a program of measurable outreach in relevant industries in order to ensure uptake. Gaps identified through this analysis should also feed into existing and future reform processes. For example, the Government can consider how potential privacy risks associated with AI could be addressed in the review of the Privacy Act 1988 currently underway.22 Guidance is also key to facilitating robust self assessment frameworks and delivering consistency as multiple industries assess and report on their use of AI technologies. DIGI highlights that these technologies will be used across the economy, not just by a small number of highly digitised companies.

The regulatory analysis previously outlined should inform sector-specific guidance that relevant regulators issue to the industries that they regulate around the application of existing laws to AI technology, noting that the context in which AI is employed is key to understanding risk. This sector-specific guidance might encourage company policies or voluntary principles to address any gaps identified where existing regulation does not address identified harms proliferated through AI. The potential risks associated with the deployment of AI vary depending on the contextual application of the technology in a variety of sectors. For example, harms that might arise from use of AI in healthcare or the financial sector will vary greatly from those presented by use of AI in entertainment, which might inform the Government’s prioritisation of resourcing regulators to undertake analysis and sector-specific regulatory guidance.

DIGI agrees with the need for risk-based frameworks for AI that take a proportionate approach to assessing risk, and that include a focus on applications that can be defined as high risk (i.e. applications whose output is likely to cause significant direct material harm). Such frameworks should provide horizontal economy-wide guidance on good AI processes, and flexibility to allow for tailored, sector- and application-specific regulation. We also believe that such frameworks have a role in promoting workable standards for explainability and transparency to promote confidence in AI technologies. In 2019, DIGI welcomed the release of the Australian Government’s AI Ethics Principles in providing guidance to a wide range of companies using AI to prevent unintended consequences and ensure the highest standards of ethical business and good governance. These principles provide a helpful framework for companies across a wide range of sectors to ensure the ethical application of AI. DIGI supports continued promotion and adoption of the AI Ethics Framework. As noted, DIGI recommends that relevant regulators issue sectoral guidance on the application of their legislation to AI-proliferated harms. This guidance should also identify gaps that might be addressed through company-level policy and industry voluntary standards; In so doing, relevant regulators should consider the relevance of the existing AI ethics principles in these recommendations to relevant industries.

In determining liability or assessment of risk related to an application of AI, DIGI recommends that, through sectoral regulatory guidance or otherwise, the Government establishes a clear delineation between organisations and developers. Various stakeholders in the supply chain play a role in the safe and responsible development and deployment of AI. Impact assessments should be present for each of these intermediaries. Assessment, monitoring, documentation and reporting should be present and proportionate to the control of the entity. For example, the organisation deploying the technology is best placed to understand the associated use cases for it and potential risks for that case. Relatedly,

22
Attorney General’s Department (2023) Privacy Act Review report, https://www.ag.gov.au/rights-and-protections/publications/privacy-act-review-report

6
developers should provide deployers with a level of transparency that facilitates accurate assessment.
While a developer cannot accurately assess the full potential risk of the technology without complete oversight of the context in which it will be deployed, safeguards can also be built into models upstream to help prevent a number of downstream risks. DIGI encourages further consultation with a range of industries, AI deployers and developers to ensure that any risk framework that is developed addresses the right risks in their sector, potential use cases, applications, and potential associated harms.

Many risks related to AI are also found in existing human systems and decisions; therefore an understanding of the most common risks at a sectoral level is conducive to effective governance, and should be the first step, prior to the consideration of how these same risks might be amplified through AI.
Noting this, DIGI believes that the application of sectoral specific regulation is best placed to ensure safe development of AI as it can best address specific use cases of the technology. This applies to both public and private sector applications of the technology. We caution against recommendations for regulatory or centralised bodies focused on reviewing the technology of AI itself, as such an approach may not give due weight or nor allow for sufficient expertise on contextual, sectoral application.

DIGI further notes that any such proposal may face significant human resources challenges, as an extremely high level of both technological expertise in relation to AI would be required alongside highly in-depth, sector-specific knowledge of every industry and government vertical where AI is applied. A more scalable, technology-neutral approach lies in upskilling and resourcing all relevant regulators. As the
Government considers the regulatory framework for AI, DIGI recommends that it consider nominating a national agency, such as the CSIRO, to offer technical advice to the relevant sectoral regulators for specific applications of the technology. Appointing the CSIRO to consult on existing regulatory frameworks and make recommendations on any updates that might be required will offer a more effective and streamlined approach than the establishment of a new regulatory or oversight body with a remit for AI as a technology. Harnessing the knowledge of an existing expert organisation will also provide a more rapid response in line with the pace of the technology’s development.

A focus on Australia's economic competitive advantage, international interoperability and coordination.

Clarity, coordination and cohesion across Australia’s AI governance

DIGI recommends that the Department ensures consistency and coordination in AI governance through a whole of Government response and clear guidance for regulators to reduce undue burden on industry and prevent barriers to innovation.

Regulatory settings and reform proposals across various regulators and Departments have direct implications on Australia’s competitive advantage and economic prosperity. As previously noted, there are significant social and economic benefits in the adoption and development of AI that the Government should be cautious of curtailing. The Government should also consider the risks present in not fostering local innovation or development of AI technologies and the broad economic impacts including to
Australia’s long term prosperity and productivity. To best support innovative and responsible development of AI in Australia and effective safeguards for its deployment, DIGI recommends that there is clarity and consistency in how domestic regulators define and address risk related to the technology. This will reduce any potential uncertainty or undue burden on industry that might be result from separate assessment and compliance requirements across different regulators. DIGI believes that first assessing the applicability of existing legal frameworks to the use of AI and producing detailed, sector-specific, regulatory guidance, including on how organisations can adequately demonstrate compliance, supports business conditions.

7
DIGI recommends the Government also considers how existing legislative settings might be best updated to allow for the innovative development of AI for the benefit of Australians. For example, Australia has a handful of restricted purpose based fair use exceptions for copyright infringement which, in our view, would benefit from adjustment to ensure that Australia can take advantage of the potential benefit of AI technologies. This is because the training and development of AI models often requires the input of copyrighted materials. By way of example, digital service providers are investing in research and development of innovative AI applications that can assist in the detection, removal and reporting of seriously harmful and illegal materials online, including pro-terror and child sexual abuse materials.
Companies investing in these solutions need to be able to process large volumes of illegal materials, but also ‘safe’ legal materials so that the technology can learn to distinguish between the two. However, it is not clear to what extent the existing fair dealing exceptions in Australian law for private use would enable research and development of this nature. Rather than expanding exceptions on a piecemeal basis that may be difficult to keep up to date given the pace of digital innovation, we recommend considering a flexible exception that can fairly balance rightsholder and user interests as technologies evolve.
Challenging questions also arise regarding the copyright in AI generated innovations. The Copyright Act
1968(Cth) does not expressly deal with the issue of the ownership of computer-generated works, although it is clear that there can be no copyright protection afforded to works under Australian law absent a human author.23 It is currently unclear whether works that are created by an AI program may therefore not benefit from copyright protection. DIGI believes this is a clear example of how existing frameworks could and should be clarified and updated to address AI, in order to ensure that Australians benefit fully from the social and economic impact of the technology’s applications.

Australia’s governance response and international models
DIGI recommends careful consideration of international interoperability and coordination on AI governance. Both development and governance of AI is in flux at a global scale. Recent research by
Resolve Strategic on Australians’ attitudes to AI indicated an understanding that the technology is relevant to Australia’s role in a global setting. The survey found that a majority (44 percent) believe AI should be regulated internationally rather than through a separate domestic approach. The results also considered options for domestic regulation alone or no regulation.24 There are many initiatives currently in development to introduce international standards, such as those produced by the ISO25, or novel regulation, such as the EU AI Act.26 DIGI encourages the Government to consider global approaches, such as OECD initiatives or the Global Partnership on AI. We believe there is benefit in identifying Australian expert stakeholders to represent domestic interests in these processes. This will ensure that Australia remains in step with international standards and is best placed to participate in the global ecosystem.
However, DIGI recommends that the Government remains cautious of replicating any one international model until its efficacy can be better assessed.

DIGI has cautioned against the establishment of a central AI regulatory authority. Similarly, we caution against the introduction of broad legislation aimed at regulating the technology and science of AI itself.
Looking to the EU AI Act, the proposed sweeping regulation across a broad range of uses is concerning as it is not based on clear evidence to suggest it is the most effective form of regulation. It also imposes burden on business, which will stifle innovation and impact investment toward AI in the EU market. Broad legislation focused on AI as a technology could also present potential conflicts with existing legislation, creating uncertain conditions for business while potentially subjecting organisations to duplicative regulatory and reporting processes. DIGI encourages the Government to consider further examples of phased and sectoral approaches, such as the UK’s pro-innovation model which adopts an outcomes
23
Telstra Corp Ltd v Phone Directories Co Pty Ltd [ 2010] FCAFC 149
24
‘Resolve Strategic’ poll conducted in June 2023
25
ISO (2022, Framework for Artificial Intelligence (AI) Systems Using Machine Learning (ML),
https://www.iso.org/standard/74438.html
26
EU AI Act 2023, https://artificialintelligenceact.eu/

8
based approach and first considers updates to existing regulatory schemes.27 At a broader scale, DIGI recommends a cohesive approach to digital regulation that takes into account the existing regulatory scheme and suggests the Government considers overarching principles to guide technology policy development, such as the principles seen in the UK Digital Regulation Plan.28

The Australian Government should consider the need for flexibility and adaptability in any regulatory framework given AI’s rapid development. There is an opportunity to consider international coordination on voluntary commitments, such as multi-stakeholder partnerships in relation to watermarking and user notice on AI generated content. At the recent White House Summit, technology companies made voluntary commitments under three key principles to guide AI governance: safety, security, trust.29 These commitments range from testing protocols to user notice initiatives to help consumers better understand when they are interacting with an AI system or AI generated content. This approach highlights the commitments already underway by AI developers at a global scale to ensure the responsible development of the technology. It also demonstrates how industry and governments can collaborate to best determine governance models for AI as the technology develops by harnessing the technical expertise and deep understanding of potential use cases. Voluntary commitments allow speed and flexibility to adapt in a shifting technological ecosystem. Exploration of internationally coordinated, voluntary principles and frameworks will allow for a rapid and internationally coordinated approach to AI governance. It also has the potential to strengthen economic opportunities at a global scale by ensuring Australia does not lag behind or diverge from the approaches of its major trade partners.

AI technologies are being developed at a global scale. There is immense complexity and cross-economy implications to be considered in the implementation of any governance framework. Regulatory developments should carefully consider international partnership and relationships. The Government should consider international economic implications, including the competitiveness of Australian business and research capabilities, the ease of conducting international business, and implications with
Australia’s international trade partners. A heavy focus on domestic regulation without a clear consideration of international cooperation and harmonisation across multiple jurisdictions could create conditions where economic opportunities are missed without a meaningful reduction in risk. Australia’s approach to international collaboration should also be considered more broadly, for example through partnerships that improve security or innovation by enabling trusted data flows across national borders for training purposes or best practice alignment on security practices.

International coordination and consultation will be essential to Australia’s role in the global AI ecosystem and ensuring best practices in governance. However, DIGI cautions against focusing solely on emerging regulatory developments in overseas jurisdictions to define domestic regulation, without consideration of the Australian context, particularly the potential economic opportunities and barriers previously discussed in this submission. First and foremost, it is important that the Government considers the evidence base and risk definition of appropriate local use cases for the technology and the appropriate responses within existing local legislation.

Conclusion
DIGI thanks the Department for the opportunity to provide its views on the Discussion Paper and its key themes. We applaud the Australian Government’s consultative approach and commitment to balancing the need to seize the opportunities presented by the development of AI, while ensuring safe and responsible use of the technology. DIGi believes that it is essential that technological, ethical, and legal

27
UK Government (2023), A pro-innovation approach to AI regulation, https://www.gov.uk/government/publications/ai-regulation-a-pro-innovation-approach/white-paper
28
UK Government (2022), https://www.gov.uk/government/publications/digital-regulation-driving-growth-and-unlocking-innovation/digital-regulation-dr iving-growth-and-unlocking-innovation
29
White House (2023) https://www.whitehouse.gov/wp-content/uploads/2023/07/Ensuring-Safe-Secure-and-Trustworthy-AI.pdf

9
expertise is harnessed to design the most effective governance framework in relation to AI technology and its applications. We encourage the consideration of a proportionate, risk based approach to regulating AI that carefully considers the existing regulatory settings and cohesion across the whole of the government. We look forward to further engagement with the Government on the next steps of this important process.

10