The Australian Government wants “guard rails” placed on the rollout of AI to address concerns such as AI taking people’s jobs and endangering lives in risk areas such as self-driving cars.
Science and Industry Minister Ed Husic has released the government’s interim response to the 510 submissions it received in response to the Safe and Responsible AI paper released in June last year.
Recommended measures include the watermarking of AI-generated content, the mandatory testing and labelling of AI models before they are implemented, and an expert advisory committee to assist in what is a fast-moving area of technology that outpaces regulators. The advisory committee would develop a voluntary AI safety standard to assist businesses implementing AI solutions.
According to the government’s paper, AI and automation could grow Australia’s economy by $170bn to $600bn a year. But high risk uses such as using AI to assess job applications and to operate self-driving cars needed examination. Transparency measures such as publicly reporting how an AI model was trained is also under consideration.
Mr Husic would require the developers of GPT products such as Microsoft, OpenAI and Google to comply with Australian law.
The government separately is looking at the proliferation of deep fakes online, along with breaches of copyright and privacy perpetrated by AI training models.
Its preference is to add provisions to stop the excesses of AI in existing law rather than develop separate legislation such as an AI Act in the European Union.
Observers will be watching closely whether this plan not to follow the EU’s approach to AI regulation flows through to other areas of regulating the big tech firms, especially when it comes to anti-trust laws.
The European Union through its executive European Commission has taken a long term stand against the excesses of the big tech firms: their anti-trust manoeuvrings to gain market advantage, and their disregard for the rights of consumers, especially privacy.
European users are set to benefit when its Digital Markets Act takes force in March; whether Australian users get a similar benefit through Australian law in the longer term remains to be seen.
The report’s release comes a day after Microsoft unveiled its commercial generative AI product, Copilot Pro, which includes an ability for individuals and companies to create their own version of ChatGPT.
The Tech Council of Australia said providing clarity on the Australian Government’s approach to AI regulation would help businesses better plan for building, investing in and adopting AI products and services. The public could take confidence that AI risks are being safely managed and regulated in Australia, it said.
“This approach is the best way to balance innovation with the need to ensure that AI is developed safely and responsibly”, said Tech Council CEO Kate Pounder.
“Australia has a sound, existing legal framework relevant to AI, and expert regulators that regulate products using AI, including in areas such as health and financial services,” Ms Pounder said. “The TCA welcomes the Government accepting our recommendation to undertake a gap analysis to build on this existing work as the first step to getting Australia’s regulatory framework right.”
She said the the establishment of an expert advisory body was one of the council’s key recommendations.
“We need to build our tech talent pipeline, upskill our workforce, increase investment in AI research, development, and commercialisation, provide organisations with the right AI tools and assurance frameworks, and build digital literacy across the community”.
Google yesterday announced it would give EU consumers a right not to cross-link their Google services as a privacy measure. This is part of Google’s measures to comply with the Digital Markets Act.
Copyright remains another sticking point with authors up in arms about tech firms using their works to teach generative AI models without acknowledgement or compensation.
Meta faces legal action for training its AI large language models using a repository of 196,000 pirated works including books by Australian authors.
The Australian Society of Authors says it received phone calls or emails from over 150 affected authors. Helen Garner, Richard Flanagan, Dervla McTiernan, Sophie Cunningham, Trent Dalton, Shaun Tan, Markus Zusak, Tim Winton, Christos Tsiolkas, Andy Griffiths, Melissa Lucashenko, Tom Keneally, Jane Harper, and Robbie Arnott were among those impacted.
Yesterday Salesforce CEO Marc Benioff sided with those concerned about AI’s plundering of copyrighted material.
Mr Benioff, who also owns Time Magazine, said artificial intelligence companies ripped off intellectual property to build their technology, in a report by Bloomberg.
“All the training data has been stolen,” Benioff said in a Bloomberg session at the World Economic Forum in Davos. Mr Benioff said content from media outlets including Time and the New York Times had surfaced in results from AI companies, according to the Bloomberg report.
He told the session that nobody had assessed what a fair price for this data was but AI companies should standardize payments to treat content creators fairly.
