If you’re running a small business and you use AI tools, pay attention to this one.

On Sunday, the Albanese government confirmed that its Digital Duty of Care framework will explicitly cover AI companies, including Anthropic (the company behind Claude) and OpenAI (the company behind ChatGPT). This isn’t vague regulatory talk. The Sydney Morning Herald and The Guardian both confirmed that private AI players will face formal obligations, potentially including mandatory risk assessments and public reporting on the consequences of their systems under Australian law.

Until now, when an AI tool gave a customer bad advice, produced wrong output, or leaked data, the legal exposure landed squarely on you, the business owner who chose to use it. That may be about to change.

What “duty of care” actually means

Duty of care is a legal concept most Australians know from workplace safety and medical malpractice. In plain terms, it means a party has a legal obligation to avoid acts or omissions that could reasonably harm others. Breach it, and you’re liable.

Under Labor’s proposed framework, AI companies would need to build safety has in from the start, not wait for individual complaints to roll in. Instead of a business owner reporting a problem and hoping the AI company fixes it, the AI company would be legally required to proactively identify and address potential harms before they reach users. Communications Minister Anika Wells is leading that work, according to The Guardian. Attorney General Michelle Rowland is heading up separate rules around automated decision-making by government agencies, with Centrelink named as a specific concern.

What it means for your business right now

If you use an AI tool and it causes harm to a customer — a wrong financial calculation, a misleading product description, a privacy breach — who wears the legal risk? Currently, in most cases: you do. The AI company’s terms of service almost certainly limit their liability, and Australian consumer law hasn’t yet caught up with how businesses actually use these tools.

The duty-of-care framework would push some of that accountability upstream to the AI providers. If OpenAI or Anthropic failed to adequately identify a foreseeable risk in their system, they could face legal consequences, not just you.

That doesn’t give you a free pass. You still choose which tools to use, and you’re still responsible for how you apply them. But it changes the equation. Right now, you’re assuming risk that arguably belongs partly with the companies building these systems.

What to watch for

Read your AI tool’s terms of service before they change. Once this legislation gets closer to reality, AI companies will likely update their ToS quickly. Some may narrow their liability exposure further; others may add compliance language that actually protects you. Know what the current terms say before the amendments come.

Document AI-assisted decisions. If you use an AI tool to draft contracts, generate customer advice, price products, or make any consequential business decisions, keep a record. Note which tool was used, what output it produced, and what human review happened. If something goes wrong and there’s a dispute, that paper trail matters.

Check if your tool is covered. This legislation targets large AI providers operating in Australia: OpenAI, Anthropic, Microsoft’s AI services, Google’s AI offerings. Smaller niche tools may not be in scope, at least initially. If you use a mix of AI tools, understand which ones would fall under a duty-of-care regime and which wouldn’t.

Why this is actually good news for small businesses

Small businesses rarely win against tech giants in a contract dispute. The terms of service you agreed to when you signed up for ChatGPT or Claude were written by teams of lawyers working for companies worth hundreds of billions of dollars. You had no negotiating power.

Imposing a duty of care on those companies gives Australian regulators a lever to pull that doesn’t rely on you having the resources to sue an American tech company. If an AI system causes harm at scale (and that includes harm to small businesses and their customers), the obligation to fix it sits with the provider, not just with whoever happened to be using the tool when it went wrong.

Assistant Minister Andrew Charlton put it plainly on Sky’s Sunday Agenda: “This is the most important technology of our lifetimes.” Labor wants AI investment in Australia, but it also wants rules that protect Australians. The duty-of-care framework is one attempt to square that circle.

Legislation is expected from next year. Watch the Legal AI hub for updates as the details firm up.

Sources