Claude Is Watermarking AI Content. Why Should Australian Businesses Care?

Anthropic has started introducing machine-readable watermarks into content generated by supported Claude models.

The immediate discussion is about detection: can people tell whether something was created with AI?

For Australian businesses, the more important question is:

Because in this case, it isn’t Australia.
Anthropic’s changes are linked to European AI transparency requirements, but they apply globally. An Australian organisation can therefore inherit a standard shaped overseas simply by using the same technology.

That matters because Australia is still working out its own position.

South Australia has announced Australia’s first Royal Commission into AI, starting 1 October 2026, while the Federal Government is also developing new AI standards through the Office of AI.

So while Australia is still debating the settings, some of them are already arriving through the technology itself.

This is where the issue becomes more practical.
A document touched by Claude could have been:

Those are very different uses.

So the real governance question is not: Was AI used?

It is:

What was AI used to do, and was that appropriate?
What does good AI governance look like?

It does not need to mean a large policy or approval for every prompt. At a basic level, an organisation should know:

What AI are we using?
What is it allowed to do?
What information can it access?
Where does human judgement sit?
Who remains accountable?

A simple way to think about risk is:

Assist → Draft → Recommend → Decide → Act

The further AI moves from assisting a person towards making decisions or taking action, the stronger the controls should become.
Using AI to summarise a meeting is different from using it to draft client advice.
Drafting advice is different from recommending which candidate to hire.
And recommending is different again from allowing an AI agent to make the decision itself.
The controls should follow the consequence of the use case, not simply the name of the product.
Saying “Claude is approved” or “Copilot is approved” is not enough.

The same tool can be used for something low risk or something that carries significant business consequence.

Unsure where your organisation stands? Take SMC’s Data and AI Readiness Assessment to review the foundations for using AI responsibly.

For many organisations, AI governance may become a commercial issue before it becomes a legal one.

• Clients may ask whether AI was used in a deliverable.
• Tenders may require disclosure.
• Contracts may restrict what information can be provided to external models.
• Boards may require higher-risk AI use cases to be registered or reviewed.
• None of that requires a new Australian AI Act.

Existing obligations around privacy, contracts, cybersecurity, consumer law and director responsibilities also continue to apply.

The practical issue is whether the organisation can explain what happened when someone asks.

Want to check where automated decisions may already be happening in your systems? Read AI Governance Is Arriving Through Existing Laws for five questions to ask your team and vendors.

The Claude watermark is useful because it shows how quickly AI standards can cross borders.
Australia is still developing its own rules.
At the same time, global providers and overseas regulators are already changing the tools Australian organisations use every day.
The response is not to stop using AI.
It is to be clearer about where it is being used, what role it is playing, what the boundaries are and who remains accountable.

If you need help putting those boundaries in place, explore SMC’s AI consulting services or book a call with Roger Molina.