The AI models behind ChatGPT, Gemini and Claude are all built by American companies. That is not the risk. The risk is where your matter data is processed, whether it is kept or used to train anything, and who inside your firm can see which answers. Here is how we handle each one.
We never put client files into the free public versions of ChatGPT, Gemini, Claude or Copilot. Those services can keep chat history, may store data overseas, may use it for training, and have no idea which of your staff is allowed to see which matter. The Victorian Legal Services Board says the same thing.
We run the same models inside Microsoft's or Amazon's Sydney data centres, under a business contract that turns off training and data retention. Everything sits in your own account, not ours. Before a model sees a single document, we check that the person asking is allowed to open that matter. Every question and answer is logged. A lawyer approves anything before it is filed or sent.
Microsoft, Amazon, OpenAI and Anthropic are American companies and remain subject to American law even when the servers are in Sydney. Most private client firms accept this once the data stays in Australia and the contract forbids training. If you act for government or in matters where that is not acceptable, we can discuss running a smaller model entirely on your own hardware. It is slower and costs more.
You stay responsible for what your lawyers rely on. The system supports your confidentiality obligations. It does not take them over.
This page is general information about how we work, not legal advice. Check with your own advisers and your professional indemnity insurer before any pilot.