Security News

Australian Firm Launches 'Sovereign' AI Model, Promising Local Data Hosting and Security Controls

Key Takeaway Before putting business or customer data into any AI tool, including locally hosted ones, ask the provider where data is stored and how it is protected, and do not rely on marketing claims alone.

Trellis Data has launched Agentaus 1, a 220-billion-parameter AI model it describes as designed for "sovereign" AI use by Australian businesses and government. The company says it was built with safety and security requirements in mind from the outset. The launch follows the federal government's plans for new AI safety and security protections, which were flagged earlier this week.

According to Trellis Data, the model includes "Defence-grade encryption" and local data hosting. It is aimed at concerns such as control of sensitive data and rising technology costs as organisations pursue AI-driven productivity gains. The company says it has a Canberra-based team supporting deployments, and that its background includes work in government, intelligence, defence and law enforcement environments. Agentaus 1 is available through the Agentaus platform on free, subscription, enterprise and API tiers.

The company says the model is about 50 per cent larger than its previous one and cites a "world ranking of 43" on the MMLU-Pro benchmark. It also claims improvements in accuracy, multi-step reasoning and output length, and says the model can run at scale on smaller compute clusters. Founder Michael Gately said the product already meets the standards the government is seeking to make mandatory. These are the company's own claims, and the source does not report any independent verification.

AI security data sovereignty Australia AI regulation data protection
Wondering what the rules are? Australia's AI guardrails in 2026, in plain language ->

Summarised by CISO AI from Australian Cyber Security Magazine, written with Claude Sonnet 5.5. We link back to every original so you can read it yourself.