In brief
2023 has been a big year in Australia for developments in artificial intelligence (AI). We've pulled together the main announcements, key insights and regulatory themes to emerge this year and recommendations for risk management that will be of interest to Australian companies engaging with AI technologies into 2024 and beyond.
If this article is of interest, you may also enjoy reading our 2023 End of Year Wrap Up on Cybersecurity in Australia.
Contents
Overarching Themes and Recommendations for AI Risk Management
In more detail: Key Developments of 2023:
1. Chief Scientist's Report on Generative AI (March 2023) ("Gen AI Report")
2. Australian Government's AI funding announcement (May 2023)
3. CSIRO NAIC report: Implementing Australia's AI Ethics Principles (June 2023) ("RAI Report")
4. Safe and Responsible AI in Australia Discussion Paper and Consultation (June - Sep 2023)
5. The eSafety Commissioner's Position Statement on Generative AI (August 2023)
6. Establishment of the Artificial Intelligence in Government Taskforce (AIGT) (September 2023)
7. AI Safety Summit and signing of the Bletchley Declaration ("Declaration") (November 2023)
8. Copyright reforms: Government establishes copyright and AI reference group (December 2023)
Setting the Scene
With Australia's AI month drawing to a close on 15 December 2023 amidst ongoing debate about a national AI strategy, regulatory framework and AI Commission, and alongside significant US and EU initiatives such as the Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence and Europe's historic deal on the Artificial Intelligence Act, we take this opportunity to reflect on some of the significant Australian policy and regulatory developments relating to Artificial Intelligence (AI) technologies that have emerged throughout 2023.
Among many highlights, 2023 saw Australia commit to implementing Responsible AI in industry, the military and the public sector, culminating in Australia signing the 2023 Bletchley Declaration on AI Safety at the UK AI Summit signalling Australia's ongoing commitment to work with the international community to ensure AI is developed with appropriate guardrails in place.
Earlier in the year, in February, Australia's National Science and Technology Council commissioned a Rapid Response Information Report on Generative AI: Language models and multimodal foundation models. The ensuing Chief Scientist's report, released in March, looked into opportunities and risks of applying AI technologies in the years to come as well as strategies that have been put in place internationally to address the potential opportunities and impacts of AI.
As part of May's 2023-24 Budget, the Australian Government announced A$101.2 million in funding to support businesses to integrate quantum and artificial intelligence technologies into their operations, affirming the Government's intent to ensure that policy and regulatory reforms will support local AI innovation. The Government's Mid-Year Economic and Fiscal Outlook (MYEFO) released on 13 December also prioritises themes of sustainability, digitisation, cyber security and digital skills.
In June, the CSIRO's National Artificial Intelligence Centre (NAIC) released its report on Implementing Australia's AI Ethics Principles with guidance on a selection of Responsible AI practices and resources to help bridge the gap between the Australian AI Ethics Principles and the business practice of responsible artificial intelligence ( RAI). Also in June, the Department of Industry, Science and Resources released a discussion paper (Safe and Responsible AI in Australia) and opened a consultation seeking views on how the Australian Government can mitigate any potential risks of AI and support safe and responsible AI practices, stimulating wide-ranging debate about the future of AI regulation in Australia and attracting 510 submissions from a full cross-section of government and industry organisations, with 448 of those responses published online.
In August, Australia's eSafety Commissioner published its Position Statement on Generative AI which evaluates the existing landscape of generative AI, the technology's life cycle and examples of positive and negative uses of it that will inform the eSafety Commissioner's regulatory approach to Generative AI. The Digital Transformation Agency (DTA) and the Department of Industry, Science and Resources (DISR) then established the Artificial Intelligence in Government Taskforce (AIGT) in September which aims to develop an integrated regulatory framework for safe, ethical, and responsible AI use across the Australian Public Service (APS).
At the AI Safety Summit hosted by the UK Government in November, Australia along with the EU and 27 countries including the US, UK and China, signed the Bletchley Declaration affirming that AI should be designed, developed, deployed, and used in a manner that is safe, human-centric, trustworthy and responsible. Earlier this month in December, the Australian Government announced it is establishing a copyright and artificial intelligence (AI) reference group to better prepare for future copyright challenges emerging from AI.
Please read on as we summarise the key highlights, themes and practical risk management recommendations to emerge from these AI regulatory and policy developments that should be of interest to all Australian organisations engaging with AI technologies into 2024 and beyond
Overarching Themes and Recommendations for AI Risk Management
2023 has seen a significant uplift in investment in and implementation of both generative AI and conventional AI technologies across a range of applications and industry settings. As organisations in Australia and around the world seek to leverage the power of AI, geopolitical competition for generative AI resources and capabilities is also increasing and there is a trend towards the concentration of generative AI resources in a handful of countries (such as the United States, China, UK, and India). For Australia, this increased competition could present challenges in terms of access to effective generative AI technologies and capabilities, as well as constraints in terms of the suitability of externally built AI models for local contexts and needs.
Along with great opportunities, the adoption of AI technologies also presents risks that need to be effectively managed across a range of areas, including:
| • | technical system risks – relating to lawfulness of data and information used to train generative AI, cybersecurity concerns and issues around the accuracy, quality and ownership of AI generated outputs; |
| • | contextual and social risks – relating to transparency, use of AI in law enforcement and health contexts and issues of representational bias that may contribute to reinforcing social inequalities and biases; |
| • | systemic social and economic risks – impacts on environment, democratic systems and social discourse (through the use of generative AI to spread misinformation and disinformation), and employment (due to job losses from automation of human tasks); and |
| • | a wide range of specific online safety risks identified by the eSafety Commissioner, including the potential for generative AI to produce Child Sexual Exploitation and Abuse (CSEA) Material and non-consensual imagery (such as deepfakes), to be leveraged by terrorist groups to disseminate pro-terror content and misinformation, and to manipulate and abuse individuals through convincingly impersonating human conversation, enabling automated personalized hate speech, bullying, abuse, and manipulation at scale. |
In this context, feedback from a variety of stakeholders in response to recent consultations supports the need for enhanced government regulation and policy to manage AI risks while supporting local innovation and opportunities, centered around a risk-based approach which would depend on the use case of the AI and the impact the relevant technology may have.
Businesses engaging with AI technologies can prepare now for upcoming regulatory reform by adopting risk mitigation practices aligned with Australia's Eight AI Ethics Principles and principles of Safety by Design. This might include:
| • | designing for human autonomy, promoting user control and feedback (e.g., by enabling users to provide feedback on algorithmically derived outputs with a "thumbs-up" button); |
| • | using dataset pre-processing, algorithmic in-processing and/or decision post-processing to measure, monitor and promote fairness of outcomes achieved using AI technology; |
| • | before launching an AI system, consulting privacy and security experts, and asking questions like "should we?" and "are you comfortable explaining it to a family member?"; |
| • | promoting reliability and safety by curating datasets, conducting pilot studies in a controlled environment to discover problems, iterate and scale the AI system; and |
| • | promoting transparency and explainability by making adequate disclosures, documenting key system information and explaining the system to the full range of potential audiences. |
In more detail: Key AI Developments of 2023
1. Chief Scientist's Report on Generative AI (March 2023) ("Gen AI Report")
Generative AI is a subset of AI technology that has attracted a lot of interest and investment in Australia and around the world throughout 2023 because it presents a broad spectrum of applications and potential for rapid integration across various industries including health, engineering, social services and creative sectors.
At the same time, with the rapid expansion and implementation of generative AI solutions such as ChatGPT across a range of industries, it is difficult to accurately predict the potential impacts of the large-scale adoption of generative AI we are currently witnessing. This poses challenges for governments and regulators seeking to manage risks associated with generative AI without stifling local innovation or depriving local organisations of opportunities.
In this context, the Gen AI Report starts off by drawing an important distinction between "Conventional AI" and "Generative AI". Generative AI is capable of categorising or identifying features of input and generating novel content in response to user prompts as either text (Large Language Models) or other media (Multimodal Foundation Models). Alternatively, Automated Decision Making (ADM) uses conventional AI to assist humans in analysing data to decide certain outcomes. Both forms of AI have a role to play in automating business processes to deliver efficiency gains, however generative AI is arguably the more controversial and revolutionary of the two because of its potential to create novel work product on a relatively independent basis and replicate creative processes that have, until recently, only been able to be performed by humans.
Key risks for businesses to consider
The Gen AI Report offers insights into the foreseeable risks and opportunities associated with generative AI based on current adoption trends. In particular, there are three crucial categories of risk that organisations should consider when evaluating opportunities in relation to generative AI, which are also likely to shape future regulatory reforms:
| • | Technical System Risks: these cover the lawfulness of data used to train generative AI, including challenges surrounding privacy, consent and copyright in the collection and use of datasets. Cybersecurity concerns relating to the potential for cyberattacks seeking to extract training data containing personal and sensitive information are also discussed. Issues around the accuracy, quality and ownership of AI generated outputs are also identified. |
| • | Contextual and Social Risks: these encompass risks to human rights and values, especially concerning the use and transparency of generative AI models, which operate as "black box technology" in high-risk contexts (such as law enforcement and health). Issues of representational bias arising from training generative AI models on specific datasets, contributing to social inequalities and biases in areas like medical research, are also discussed. |
| • | Systemic Social and Economic Risk: these include impacts on environment (relating to the significant energy and water consumption of training and operating generative AI), democratic systems and social discourse (through the use of generative AI to spread misinformation and disinformation), and employment (due to job losses from automation of human tasks). |
The Gen AI Report also identifies risks associated with increased geopolitical competition for generative AI resources and capabilities, amidst a trend towards the concentration of generative AI resources in a handful of countries (such as the United States, China, UK, and India). For Australia, this increased competition could present challenges in terms of access to effective generative AI technologies and capabilities, as well as constraints in terms of the suitability of externally built AI models for local contexts and needs.
Despite some government investments in AI-related initiatives to date, requirements for skilled workers and significant computing and data requirements present challenges that may constrain Australia's ambitions to be a leader in the development and implementation of generative AI technologies.
The Gen AI Report also highlights a lack of specific regulatory frameworks addressing generative AI around the world. Various regulatory frameworks addressing AI generally are in development, including the European Union's proposed AI Act, which is notable for differentiating between AI use cases, banning unacceptable uses and categorising some as "high risk."
Why does it matter?
The risks and opportunities identified in the Gen AI Report will influence the Australian Government's approach to policy and regulation of generative AI as it seeks to manage risks while promoting local innovation and opportunities associated with generative AI.
2. Australian Government's AI funding announcement (May 2023)
As part of the 2023-24 Budget, the Australian Government announced AUD$101.2 million in funding to support businesses to integrate quantum and artificial intelligence technologies into their operations.
This builds on the AUD$1 billion set aside for critical technologies through the National Reconstruction Fund and includes:
| • | $40.2 million to deliver a Critical Technologies Challenge Program focusing on projects that use quantum computing; |
| • | $19.8 million to establish the Australian Centre for Quantum Growth to support research and development of a quantum technology industry in Australia; and |
| • | $41.2 million to support the responsible deployment of AI in the national economy, with some funding set aside for grants to support small to medium enterprises without AI expertise to adopt AI technologies. |
Why does it matter?
This funding commitment demonstrates the Government's recognition that AI and quantum computing present great economic opportunities for Australian industry, as well as its intent to ensure that policy and regulatory reforms will support local innovation. However as noted above, the constraints of intense geopolitical competition for generative AI resources and capabilities and the need for skilled workers and substantial computing and data resources mean that significant ongoing investment will be required to support Australia's ambitions to be a leader in the development and implementation of AI technologies.
3. CSIRO NAIC report: Implementing Australia's AI Ethics Principles (June 2023) ("RAI Report")
After the 2022 Responsible AI Index found a concerning 'action gap' – between the 82% of businesses expressing a belief they are practising AI responsibly and the 24% with actual measures in place to ensure their AI systems are developed responsibly – the NAIC released the RAI Report to support businesses to mature their Responsible AI (RAI) practices, and in the process manage both legal and reputational risks associated with the use of AI.
Key risks for businesses to consider
Legal risks can arise if AI systems fail to comply with legal requirements including those relating to anti-discrimination, privacy, consumer protection and intellectual property – for example, a conventional AI system might make an inscrutable decision in a situation when the law requires that decision to be explainable or process an individual's personal information for purposes that the individual does not consent to, or a generative AI system might infringe intellectual property rights when ingesting third party content without an appropriate licence in place. Reputational risks may also arise if the use of AI in a particular context, even if legal, is considered controversial, unethical or untrustworthy by customers, regulators or the public.
Risk management recommendations aligned with AI Ethics Principles
To mitigate the identified risks, the RAI Report outlines various practical steps businesses can take to design, deploy, maintain and use AI systems in a way that is accountable to the people the AI systems interact with, minimises the risk of negative consequences and maximises the benefits to individuals and society. The RAI Report is structured around Australia's Eight AI Ethics Principles, and in total identifies 26 different pragmatic practices businesses can adopt to implement those principles along with useful resources and guidelines businesses can refer to when implementing those practices.
Practical tips for all businesses to consider when designing and implementing AI systems include:
| • | designing for human autonomy (including by establishing user consent and promoting individual user control over an algorithm, e.g., by enabling users to provide feedback on algorithmically derived outputs such as with a thumbs-up button found in many recommender systems such as online music and video streaming platforms); |
| • | justifying the means by which positive impact is achieved (before launching an AI system, asking questions like "should we?" and "would you be comfortable explaining the ideas behind the system to a friend or family member?"); |
| • | incorporating human diversity into the design of the system (AI systems designed without diverse and inclusive practices are more likely to be perceived as untrustworthy, unfair or actively discriminatory); |
| • | defining how to measure and monitor fairness using data (fairness problems with AI systems tend to be accidental, and a system’s operators are usually unaware of them so it is important to select appropriate fairness metrics that can be used to quantify the fairness of outcomes, so that any unfair impacts can be identified, measured, mitigated and corrected or prevented going forward). Mitigations may take the form of dataset pre-processing, algorithmic in-processing or decision post-processing; |
| • | promoting privacy and security by consulting privacy and security experts before launching an AI system, minimising collection of personal information, considering privacy preserving models such as training the system on aggregated or synthetic data, and building security measures into the system focused on protecting the confidentiality, integrity and availability of data; |
| • | promoting reliability and safety by curating datasets, conducting pilot studies in a controlled environment to discover problems, iterate and scale the AI system, and monitoring and continuously evaluating the system's data, models and performance – including for issues relating to dataset shift, adversarial attacks, underperforming system infrastructure, data outliers and concept drift; and |
| • | promoting transparency and explainability by making adequate disclosures, documenting key system information and explaining the system appropriately to a range of audiences taking into consideration all stakeholders, including people the system makes decisions about, people who act on the system's advice or outputs, system owners and the development team, external reviewers such as regulators and the general public. |
Why does it matter?
Applying the practical guidance in the RAI Report will help Australian businesses to comply with Australia's AI Ethics Principles. This is important to manage risks under existing legal frameworks, and may also serve to help businesses future-proof AI systems in advance of anticipated upcoming regulatory reforms.
4. Safe and Responsible AI in Australia Discussion Paper and Consultation (June – Sep 2023)
In June, the Department of Industry, Science and Resources released its 'Safe and Responsible AI' Discussion Paper (Discussion Paper) seeking submissions on actions that may be taken across the economy in Australia on AI regulation and government to mitigate the potential risks of AI and to increase public trust and confidence in the development and use of AI.
The Discussion Paper focused on governance mechanisms to ensure AI is used safely and responsibly. Reflecting on current approaches around the world to AI regulation, the Discussion Paper highlighted that in Australia AI governance presently relies on a combination of:
| • | general regulation that is technology neutral (e.g., consumer protection, online safety, criminal and privacy); |
| • | sector-specific regulation (e.g., therapeutic goods, financial services); and |
| • | voluntary or self-regulation initiatives (e.g., the AI Ethics Principles). |
Key themes in consultation submissions
Indicating the economy-wide interest in AI regulation in Australia, over 400 responses were submitted providing input on the following items (amongst others):
| • | defining the various terms used to describe AI (e.g., 'machine learning'); |
| • | potential gaps in current regulatory approaches and non-regulatory approaches; |
| • | the extent to which any governance measures in other jurisdictions are suitable and adaptable for Australia; |
| • | target areas for priority for Australia regarding regulation and governance of AI; and |
| • | the utility of using risk-based approaches to manage and safeguard individuals' rights when AI is adopted. |
Submissions from business were largely consistent:
| • | Stressing the need for uniformity when defining AI terminology. Standardised definitions will not only enable consistency but support growth in Australia. Some responses even sought to adopt international standards and norms, and only creating different standards where exception was truly necessary; |
| • | Not seeking bespoke AI laws, and instead asking that the government focus on ensuring that existing regulations remained fit for purpose as AI further develops; |
| • | Being wary of multiple regulators in Australia with legislative functions relating to AI. Businesses signalled that greater guidance and coordination will be needed by the government to ensure that there is a consistent approach; and |
| • | Supporting a risk-based approach, which would depend on the use case of the AI and the impact the relevant technology may have. |
Australia's regulators also responded and considered how existing regulatory frameworks can be utilised or strengthened, including through existing reform proposals, to provide appropriate safeguards for the Australian public in relation to this technology.
Why does it matter?
Seeking economy-wide engagement to shape Australia's positioning on AI governance is important to increase trust and confidence in the development and use of AI and the adoption of AI in Australia (which is lagging in comparison to many other jurisdictions around the world).
It is particularly important for Australia to get its AI regulatory settings right considering the Productivity Commission's view that AI technologies are vital enablers of productivity and one of the transformative digital technologies that could help to drive productivity growth in Australia in the years to come.
5. The eSafety Commissioner's Position Statement on Generative AI (August 2023)
Building on the risks identified in the Chief Scientist's Gen AI Report of March 2023 (see Item 1), in August Australia's independent regulator and educator for online safety, the eSafety Commissioner, released an important Position Statement on Generative AI ("Position Statement") which highlights online safety challenges presented by advancements in generative AI that all Australian businesses should consider when implementing generative AI systems.
The Position Statement expresses concerns that the rapid evolution of generative AI technologies and perceived commercial imperatives to swiftly adopt and incorporate generative AI into business operations may cause businesses engaging with the Australian market to overlook important online safety considerations, which is an area of focus for the eSafety Commissioner.
Key risks for businesses to consider
Australian businesses should consider the following key risks identified in the Position Statement when implementing generative AI systems or enabling generative AI technologies to be used by Australian users in conjunction with their services, particularly in an online environment:
| • | Creation of Child Sexual Exploitation and Abuse (CSEA) Material: Generative AI poses a risk in generating CSEA materials using photos of children sourced from social media, presenting safety challenges for parents and young individuals. Perpetrators can also exploit the language capabilities of large language models (LLMs) powered by AI to groom children in automated and targeted ways. |
| • | Exposure of Inappropriate Content to Children: Generative AI has the potential to produce age-inappropriate content, such as violent or sexually explicit material, exposing children to content not suitable for their age. |
| • | Negative Impact on Wellbeing and Safety: Young people may seek support from chatbots and conversational AI, but there is a risk that generative AI might struggle to appropriately handle disclosures and fulfill reporting obligations when children share harmful experiences. |
| • | Non-consensual Imagery: Generative AI deepfakes may be used to create and threaten to share artificial but realistic pornography featuring real adults without their consent. |
| • | Terrorism and Violent Extremism: Terrorist groups could leverage generative AI to disseminate pro-terror content, raise funds, and generate instructions on creating weapons. |
| • | Bullying, Abuse, and Hate Speech: Generative AI can be exploited to manipulate and abuse individuals through convincingly impersonating human conversation, enabling automated personalized hate speech, bullying, abuse, and manipulation at scale. |
| • | Bias and Inclusivity Issues: Current generative AI systems may reflect biases present in publicly available online data, potentially misrepresenting diverse values and perspectives. Outputs may influence user views, values, and experiences, potentially shifting societal norms around challenging topics. |
| • | Competition and Consumer Concerns: Users may not always be aware of generative AI's use for advertising or providing information that influences their online activities and purchasing decisions. Firms may tie generative AI usage to other services, potentially creating competition issues. |
| • | Privacy Concerns: Outputs from generative AI models may contain personal and sensitive information, raising the risk of data breaches and potential harm to individuals. |
Risk management recommendations: Safety by Design
To address the wide-ranging online safety-related risks identified above, companies are urged to enhance safety measures relating to generative AI technologies by embedding three core Safety by Design principles into their AI governance processes:
| • | Service Provider Responsibility: Designate individuals accountable for user safety, conduct risk assessments, embed watermarks in AI-generated content, and establish responsive systems for user safety concerns. |
| • | User Empowerment and Autonomy: Define user rights, responsibilities, and safety expectations, offer disclaimers and content warnings for AI-generated content, implement measures like informed consent, and enable user feedback. |
| • | Transparency and Accountability: Share information on AI models and design, provide accessible safety policy details, invest in technologies for user safety, publish transparency reports on abuse and metrics, and document AI model capabilities, limitations, intended uses, and prohibitive applications. |
Why does it matter?
The Position Statement helps businesses understand the wide range of online safety issues presented by generative AI technologies and the regulatory and enforcement approach the eSafety Commissioner is likely to take in relation to those issues. It also enables businesses to prepare now to mitigate risks associated with possible regulatory scrutiny by the eSafety Commissioner in relation to their implementation of generative AI technologies, by taking steps to incorporate Safety by Design principles when implementing or engaging with generative AI technologies, and documenting how these principles are applied. Adopting a Safety by Design approach will help organisations to minimise harm to users and in the process mitigate the associated risks of regu latory scrutiny and legal liability.
6. Establishment of the Artificial Intelligence in Government Taskforce (AIGT) (September 2023)
On 20 September 2023, the Australian Government announced the formation of the Artificial Intelligence in Government Taskforce (AIGT), jointly led by the Digital Transformation Agency (DTA) and the Department of Industry, Science and Resources (DISR). Comprised of 18 representatives from 11 Australian Public Service (APS) agencies, the taskforce will focus on ensuring the safe and responsible use of AI within the APS. Over a six-month period, it will address whole-of-government AI applications, policies, standards, and guidance, consulting widely across APS to develop governance, risk management, skills, capability, technical use, and preparedness approaches.
The taskforce aims to update the guidance on government use of publicly available generative AI platforms, reflecting the rapidly evolving landscape of AI technology. It responds to the need for a cohesive, government-wide approach to harnessing AI opportunities while effectively managing associated risks.
Why does it matter?
The initiative aligns with existing government efforts, including DISR's Safe and Responsible AI in Australia discussion paper and consultation (see Item 4), and underlines the importance that the Australian Government sees in setting an example for the safe and ethical use of AI technologies to benefit the wider Australian community.
7. AI Safety Summit and signing of the Bletchley Declaration ("Declaration") (November 2023)
In November, at the AI Safety Summit hosted by the UK Government, Australia along with the EU and 27 countries including the US, UK and China, signed the 2023 Bletchley Declaration on AI Safety affirming that AI should be designed, developed, deployed, and used in a manner that is safe, human-centric, trustworthy and responsible. From an Australian perspective, signing the Declaration is a significant international development that signals Australia's ongoing commitment to work with the internati onal community to ensure AI is developed with the right guardrails in place, and acknowledges that in today's interconnected world, many risks arising from AI are international in nature and best addressed through international cooperation.
The Declaration places a particular emphasis on frontier AI capabilities and strong expectations on developers to ensure the safety of their AI systems, including through safety testing and thorough evaluations. In particular, it strongly encourages all relevant actors to "provide context-appropriate transparency and accountability on their plans to measure, monitor and mitigate potentially harmful capabilities and the associated effects that may emerge, in particular to prevent misuse and issues of control, and the amplification of other risks".
Against this backdrop, signatories to the Declaration have resolved to support:
| • | "an internationally inclusive network of scientific research on frontier AI safety to facilitate… policy making and the public good"; and | |
| • | an "agenda for addressing frontier AI risk [that] will focus on: | |
| • | identifying AI safety risks of shared concern, building a shared scientific and evidence-based understanding of these risks, and sustaining that understanding as capabilities continue to increase, in the context of a wider global approach to understanding the impact of AI in our societies; and | |
| • | building respective risk-based policies across our countries to ensure safety in light of such risks, collaborating as appropriate while recognising our approaches may differ based on national circumstances and applicable legal frameworks. This includes, alongside increased transparency by private actors developing frontier AI capabilities, appropriate evaluation metrics, tools for safety testing, and developing relevant public sector capability and scientific research." |
Why does it matter?
The commitments made in the Declaration represent a collective aspiration for AI governance across the 29 participating countries. This is perhaps unlikely to be the first step towards a unified global regulatory framework for AI as nations seem poised to adopt diverse regulatory strategies, leading to a complex tapestry of local AI-related laws, regulations and standards for international businesses to navigate.
For participants like Australia, signing the Declaration is likely to drive further domestic regulatory and policy reform in relation to AI, perhaps drawing some inspiration from recent regulatory developments in other jurisdictions that are signatories to the Declaration such as the Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence in the United States and proposals to regulate AI in Europe and the UK which take diverging approaches while maintaining in-principle alignment with the intent of Declaration.
8. Copyright reforms: Government establishes copyright and AI reference group (December 2023)
On 5 December 2023, the Australian Government announced it is establishing a copyright and artificial intelligence (AI) reference group ("Reference Group") to better prepare for future copyright challenges emerging from AI.
The Reference Group is an outcome of a series of four roundtables with key stakeholders on copyright priorities and emerging issues held in February, June, August and December 2023, which discussed issues including:
| • | the implications of artificial Intelligence (AI) for copyright law from the varying perspectives of copyright owners/creators, copyright users/consumers, the broadcasting and media sector and the technology sector; |
| • | the definition of 'broadcast' for the purposes of the Copyright Act 1968 (the Copyright Act); |
| • | a limited liability scheme for the use of orphan works (that is, works for which the copyright owner cannot be found); |
| • | the use of copyright material in remote learning environments (with participants agreeing to the overarching principle that there should be equivalence between how the Copyright Act treats physical and virtual classrooms); and |
| • | quotation from copyright material – including whether there are additional publicly beneficial purposes (beyond those already permitted under the Copyright Act) for which it would be reasonable to allow the quotation of copyright material if it is otherwise 'fair' and 'proportional', as well as additional safeguards, if any, that may reasonably be incorporated in any fair dealing reforms to provide protection for rights holders. |
Why does it matter?
The issue of AI works potentially infringing copyright is still a live one in Australia that affects stakeholders across a wide range of sectors, including the creative, media and technology sectors.
The Reference Group will engage with stakeholders across these sectors and will influence the government's approach to copyright policy and regulatory reforms affecting AI technologies, with a focus on a number of important copyright issues relating to AI, including the material used to train AI models, transparency of inputs and outputs, the use of AI to create imitative works, and whether and when AI-generated works should receive copyright protection.
Where to from here?
As 2023 draws to a close, the potential for AI technologies to improve Australian lives in meaningful ways and to boost economic productivity in Australia is clearer than ever, however it is equally clear that AI presents unique risks. AI technologies can cause significant harm if not developed and deployed safely, and opportunities may be lost if technological development and innovation are not supported by appropriate regulatory settings.
Currently Australia's regulatory environment for AI is patchwork at best and stakeholders are increasingly questioning whether this is sufficient or appropriate to address emerging risks while supporting innovation.
While Australia does have an Artificial Intelligence Ethics Framework consisting of 8 Artificial Intelligence (AI) Ethics Principles designed to ensure AI is safe, secure and reliable, this is a voluntary, principles-based framework which, although generally sound, is not in a form that can be directly legislated.
Going forward, and perhaps in contrast to the European approach, we expect the Australian Government's approach to AI regulation to be in the form of "evolution, not revolution". That is to say, we expect to see a targeted and nuanced approach to regulatory reform, commencing with detailed reviews of existing legislation to identify specific amendments (which may draw on elements of the 8 AI Ethics Principles and/or principles of Safety by Design) that are needed to address risks posed by AI technologies in particular legislative contexts, rather than developing a new comprehensive AI law – the likes of which we expect will only be introduced in the longer term if legislative gaps are identified which cannot be adequately addressed by targeted amendments to existing legislation.
To support the law reform process, the Australian Government may adopt a similar approach to the UK Government and publish a national strategy (perhaps in a similar form to Australia's Cyber Security Strategy) for how it will regulate AI, clarifying the ways in which existing laws already regulate and apply to AI technologies and how any proposed law reforms will further protect Australians from emerging risks while simultaneously supporting innovation and economic opportunities.
Whether Australia will also usher in a new dedicated AI regulator such as an "AI Commissioner" in 2024 remains to be seen. However, one thing is certain: AI regulation and policy in Australia will be an area to watch in the year to come as momentum for significant AI law reform continues to build.
