When top federal government staff were getting ready to meet with OpenAI’s leading lobbyists in June last year, they steeled themselves for some big asks.
They expected the makers of ChatGPT to pitch for Australian government investment in American infrastructure and the rolling back of our stricter approach to copyright and AI regulation — or risk missing out on all the spoils of the AI boom.
Internal government documents obtained via freedom of information requests show how the government prepared and responded to OpenAI’s early lobbying efforts in Australia.
Revealed for the first time by Crikey, these engagements go beyond what was previously known about such lobbying efforts, including the existence of a two-hour meeting between OpenAI and the Office of National Intelligence (ONI).
The push by the firm behind the world’s fastest growing tech product, ChatGPT, has had mixed results so far. Some of OpenAI’s policy requests have either been ignored or rejected. Others have been adopted by the government as official policy.
A Department of Prime Minister and Cabinet (PM&C) briefing about OpenAI, obtained by Crikey, was prepared ahead of a June 3 meeting with the department’s deputy secretary Nadine Williams. (OpenAI’s meetings with high-level politicians and public servants had been previously reported on, but not in much detail).
This document warned government staff that OpenAI “may seek Australia’s co-investment in US-based infrastructure and press Australia to walk back certain proposed and existing regulatory settings, including on copyright, privacy law and accountability”.
The briefing’s “co-investment” was about OpenAI’s “OpenAI for Countries” program which, among other goals, sought to get nations to put money into its US$500 billion US data centre buildout project called Project Stargate.
OpenAI had discussed this program with Department of Industry, Science and Resources staff prior to Australia’s 2025 federal election, the briefing said.
Government coordination
A spokesperson for PM&C said that it “did not discuss Australian government co-investment with OpenAI”.
The briefing also flagged that OpenAI — which it hailed for “precipitat[ing] the mainstream spread of generative AI” but noted criticism that it had weakened its safety testing — “may characterise Australia’s proposed mandatory guardrails for AI in high-risk settings as a potential barrier to investment.”
These newly unearthed emails between high-level public servants show OpenAI’s overtures to the national intelligence apparatus included a two-hour meeting with ONI, the federal agency tasked with producing intelligence assessments for the prime minister and top ministers.
One PM&C official noted in an email: “[First Assistant Secretary] Narelle [Luchetti] is meeting with [redacted] next Tuesday, I was also advised they are meeting with ONI for 2 hours the same day.”
The deputy secretary-level PM&C meeting brought together senior officials including Williams, Luchetti, and national security division head Kendra Morony.
On OpenAI’s side, documents show it brought staff with “International” and “Asia Pacific” responsibilities, which lines up with previous reporting of OpenAI’s chief strategy officer Jason Kwon and its Asia Pacific head of public policy Sandy Kunvatanagarn going to Canberra. Its federal lobbyist, Bourke Street Advisory’s Amelia Metcalf, was also scheduled to attend.
OpenAI’s visit triggered coordination across multiple government departments. Department of Home Affairs deputy secretary and Australia’s ambassador for cyber affairs and critical technology Brendan Dowling proposed organising a joint meeting.
Department of Industry, Science and Resources head of technology and digital Anthony Murfett and Home Affairs’ first assistant secretary for counter foreign interference, cyber and technology Peter Anstee both enthusiastically agreed.
Scepticism in government
The company’s lobbying campaign also included a previously reported exclusive dinner for senior bureaucrats at The Boat House restaurant, and releasing “OpenAI’s Economic Blueprint” for Australia calling for tax breaks and reduced regulation. (The report was compiled by Mandala Partners, the consulting offshoot of a firm founded by future Assistant Minister for Science, Technology and Digital Economy Andrew Charlton, AlphaBeta).
Later in June, OpenAI signed its first-ever federal government contract, a $50,000 deal with Treasury. Documents obtained from Treasury about the contract show this was for an unspecified amount of pooled credits for ChatGPT use (as opposed to, say, a certain number of subscriptions for staff). The company has since signed a second contract with the Commonwealth Grants Commission.
It also appointed former AlphaBeta partner and Tech Council of Australia CEO Kate Pounder to head up its Australian policy work. (Pounder’s contract with OpenAI has recently finished, but she is currently in negotiations for another senior role in the company).
But further Treasury documents reveal significant scepticism within the government about the economic promises underpinning OpenAI’s pitch.
While OpenAI’s report predicted a $115 billion annual boost to the economy in five years, a briefing prepared for Treasurer Jim Chalmers ahead of a previously reported September meeting with OpenAI executives was much less confident.
This briefing, obtained by Crikey, noted that “estimates of the magnitude and timing for benefits vary widely”. The document cited academic research suggesting AI could deliver as little as a 0.1 percentage point increase in annual productivity growth, while other projections run up to 30 times higher.
The briefing also poured cold water on claims that AI was already transforming the economy.
‘Benefits… yet to emerge’
“Evidence of wider benefits/deep integration with business operations are yet to emerge,” the Treasury brief stated, adding that labour market impacts “are uncertain” and productivity gains appear “generally task-specific”.
The briefing highlighted the resource demands of AI infrastructure, noting that data centres using evaporative cooling “can consume tens of millions of litres of water each year” and that projected growth would add to “growing pressures on water supplies”.
Since then, the federal government has launched Australia’s National AI Plan. The document, released in early December, makes no commitment to invest in OpenAI’s infrastructure ambitions or provide the tax breaks outlined in the company’s Economic Blueprint.
The same week as the AI plan was announced, Chalmers, Charlton and Minister for Industry, Innovation and Science Tim Ayres all welcomed an agreement between OpenAI and publicly listed data centre company NEXTDC for a $7 billion buildout in Western Sydney, which notably did not include any financing from the government.
The government also rejected a request to review copyright settings, resisting a push by the lobby group Tech Council for Australia and OpenAI for changes that would have made it easier to use copyrighted material for AI training without authorisation.
But the national AI plan did contain a significant win for OpenAI and the AI industries: it notably excluded the mandatory guardrails for high-risk AI applications that had been previously proposed by former science minister Ed Husic in lieu of a much lighter-touch approach.
OpenAI declined to comment.
This story first appeared on Crikey. You can read the original here.
