A Melbourne company that touted itself as one of Australia’s only builders of truly “sovereign” large language models is pulling back on the rhetoric, saying the term is too defensive and outdated, as US partners and competitors circle.

The change in language comes as Maincode prepares to unveil its flagship ‘Matilda’ platform next week and after engaging its first Canberra lobbyist amid a national debate on Australia’s AI opportunity.

The company grabbed headlines in July when it announced it has the backing of billionaire co-founder Ed Craven to build large language models on local infrastructure with curated Australian training data.

Maincode chief executive Dave Lemphers cofounded the company with the billionaire founder of online casino giant Stake.com, Ed Craven. Image: LinkedIn

Maincode’s other co-founder and chief executive Dave Lemphers has repeatedly stressed the importance of sovereign models for clients like governments and researchers that can incorporate them in a portfolio approach to the US-dominated market.

But writing on LinkedIn on Saturday, Mr Lemphers said the company has “moved away from the language of sovereignty”.

“It is not that the intent was wrong (control, security, capability), but the framing has become too loaded, too defensive, and too divorced from the day-to-day reality of actually building AI systems,” Mr Lemphers wrote.

“’Sovereign AI’ has drifted into rhetoric that divides, when what we see every day inside the factory is collaboration, creativity, and optimism.”

The company is now using language like “Australian-made AI” instead and insists it is still “building real systems in Australia by Australians” that are “aligned with our needs and values”.

“And to be clear, Australian-made AI does not mean everything is built here. It means what we build is principally created, operated, and advanced in Australia, that the opportunities, jobs, and innovation originate here,” Mr Lemphers wrote.

Mr Lemphers later told InnovationAus.com that the company will still be able to honour customers’ requests to keep data and models within Australia, but is not interested in “sovereignty as a slogan”.

“We deliberately sidestep the pundit-led debate because it’s a distraction from the real work. It encourages a focus on compliance, which ringfences opportunity for large incumbents, rather than fostering a truly Australian industry,” he told InnovationAus.com

The shift opens Maincode to working with more of the global firms that power artificial intelligence with software, data and digital infrastructure.

In September, Maincode engaged Akin Agency as its first Canberra lobbyist, according to the public register.

The firm also represents other tech sector clients in Canberra and until last month had represented the sector’s industry group, the Tech Council of Australia.

Parts of the local and global tech sector, including the Tech Council, are pushing the Albanese government for support to establish more data centres in Australia and encourage uptake of artificial intelligence.

Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton flew to the US in August to meet with AI companies Anthropic, OpenAI, AWS, Grok and NVIDIA about investing in their digital infrastructure in Australia.

“We’re open to investment from a wide range of different parties, not just one party,” Dr Charlton told InnovationAus.com after the trip.

“And a number of those parties, including OpenAI, have programs where they are working with governments, working with sovereigns. And Australia is certainly looking at what options there might be for us to maximise the investments we get from a range of different players and use that to turbocharge our domestic AI sector.”

Dr Charlton also backed the Australian sector’s capability to produce foundational and niche large language models, amid doubts about how competitive it can be.

“Australia may well not produce something that competes directly with ChatGPT. But I think we should definitely have ambitions as a nation to be participating in the creation of AI tools, including models for specific applications and purposes. And including software and other application layers on top of models.”

This article has been updated to include additional comments from Maincode CEO Dave Lemphers.

Do you know more? Contact James Riley via Email.