How Australian Businesses Can Grow with AI FAQ

Introduction

Australian businesses are increasingly embracing the use of AI (Artificial Intelligence) to drive growth and improve efficiency. This FAQ guide provides valuable insights into the use of AI, specifically LLMs (Language Model Models), in business operations. It also addresses the privacy concerns associated with using LLMs and discusses the current and potential future legislation in Australia. This article aims to help Australian businesses understand the benefits and challenges of using AI and provides best practices for leveraging LLMs while protecting consumer privacy.

What are LLMs?

LLMs (Language Model Models) are a branch of AI that utilize deep learning algorithms to recognize, summarize, translate, and generate content based on extensive datasets. These models are trained with large volumes of data, enabling them to establish relationships between different words and generate quick responses to various prompts or questions. LLMs have the potential to enhance efficiency and productivity in the commercial landscape.

How are businesses using LLM AI?

The use of LLM AI is prevalent in the IT sector, with other industries following suit. IT developers employ LLMs to write software and program robots, while scientists train LLMs to understand proteins, molecules, DNA, and RNA to gain insights into living organisms. In the legal field, law firms are using LLMs to answer simple questions, conduct research, draft documents, and emails. LLMs assist lawyers in quickly resolving basic legal queries and assessing legal risks, allowing them to focus on more complex matters that provide greater value to their clients.

What are the privacy concerns of using LLMs?

While LLMs offer numerous benefits, their use raises significant privacy concerns. The collection of personal data during interactions with LLMs, coupled with their potential to facilitate cybercrime and lack of transparency, poses a threat to consumer privacy.

Collection of personal data

Users often provide personal or sensitive information to LLMs when asking questions or providing prompts. This data is used not only to answer their queries but also to train the LLM for future responses. If the recorded data contains personal or confidential information about individuals or businesses, it could be at risk of exposure in the event of a data breach. This puts sensitive business information and clients’ private information at risk, potentially resulting in legal penalties and damage to client relationships and reputation.

Cybercrime

LLMs can be exploited by cybercriminals to create malware, phishing emails, and other malicious activities. AI-powered scams leverage LLMs to impersonate individuals, manipulate data, and deceive businesses into providing personal information or transferring funds. For example, natural language processing (NLP), an AI-powered scam, generates human-like text that hackers use to create convincing phishing emails or social media posts. These emails may reference recent transactions or incorporate personal information to create a sense of urgency, enticing victims to click on malicious links or disclose personal information. Deepfakes, which are indistinguishable from reality, can also be created using LLMs to obtain personal data or spread fake material.

Transparency

The transparency of user anonymity when using LLMs is another significant concern. It is unclear whether a user’s questions and prompts can be traced back to them, either by the organization or the LLM itself, and whether this constitutes the sharing of personal information. Some LLMs use unique identifiers and login trails to identify users and link them to their prompts or questions. This raises questions about the retention of potentially sensitive data and erodes user trust, as individuals may feel that their personal information has been mishandled or misused.

How does Australian law address these concerns?

Australia does not have specific legislation addressing AI. Instead, the use of AI is governed by existing laws. The Privacy Act 1988 (Cth) applies to the collection of personal information, and the Copyright Act 1968 (Cth) regulates the use of intellectual property rights in training LLMs and generating LLM-generated content. The Office of the Australian Information Commissioner has penalized AI companies in the past for breaching the Privacy Act, but these measures are often reactive and occur after privacy infringements have already taken place.

The Australian Government has shown a willingness to amend the current legislative landscape. For example, the Department of the Prime Minister and Cabinet released an issues paper in March 2022, seeking input on how to regulate the responsible use of AI and automated decision-making. The paper refers to the Australian Human Rights Commission’s Human Rights and Technology Final Report, which recommends the establishment of a dedicated AI safety commissioner. The paper also mentions the Review of the Privacy Act Report, which suggests greater regulation of automated decision-making and increased transparency in privacy policies.

The Australian Government’s AI Action Plan (2021) aims to position Australia as a leader in adopting trusted, secure, and responsible AI. While an official response to the submissions has not been released, it is evident that action is needed to effectively regulate the use of AI and address privacy concerns.

Potential legislative reform

To address the growing power of LLMs and AI, Australia could implement significant legislative reforms. Taking cues from other countries such as the EU and China, Australia could develop comprehensive regulations to ensure the responsible use of AI.

The EU is set to finalize the first Legal Framework on AI, known as the EU AI Act, in 2023. This legislation provides a broad definition of AI and prohibits the use of AI systems that present an “unacceptable risk.” Violating this legislation can result in fines of up to €30,000. The EU AI Act covers AI systems that may cause psychological harm, exploit vulnerabilities of specific groups, or materially distort behavior.

China has issued a Code of Ethics for New-generation Artificial Intelligence, which integrates ethics and morals into AI development. The code emphasizes improving human well-being, promoting fairness and justice, protecting privacy and security, ensuring controllability and credibility, strengthening responsibility, and enhancing ethical literacy.

Australian businesses should monitor these global developments and advocate for legislative reform that aligns with international standards.

Best practices for Australian businesses – how should businesses use LLMs?

While awaiting legislative reform in Australia, businesses can adopt best practices to utilize LLMs effectively while safeguarding privacy.

1. Implement data redaction procedures: Redact data and convert it into unintelligible forms to ensure that no sensitive data is collected by LLMs and used to inform other responses.

2. Use synthetic PII replacements: When feeding LLMs personally identifiable information (PII), opt for synthetic PII replacements instead. These replacements use contextually correct fake data to provide data security without compromising commercial use or output utility.

3. Consider private LLMs: Instead of using publicly available LLMs, consider utilizing private programs that protect PII within text inputs and only share necessary and synthetic information with other language models. Private LLMs allow businesses to leverage the advantages of LLMs while safeguarding customer privacy.

4. Backup data regularly: Given the increased instances of cybercrime facilitated by LLMs, businesses should consistently back up their data to mitigate the impact of any data breaches. Regularly assess the adequacy of data handling procedures to ensure data security.

Conclusion

Australian businesses can harness the power of AI, particularly LLMs, to drive growth and improve operational efficiency. However, privacy concerns associated with LLMs must be addressed. While legislative reform in Australia is pending, businesses can adopt best practices to protect consumer privacy while leveraging the benefits of LLMs. By implementing data redaction procedures, using synthetic PII replacements, considering private LLMs, and regularly backing up data, businesses can navigate the challenges associated with LLMs effectively. Monitoring global developments and advocating for legislative reform will also help shape a regulatory framework that balances the benefits of AI with privacy expectations.