AI has given many Australian business owners a tempting story: “We have built our own proprietary tool.” The test is whether that tool would survive due diligence, or whether part of it quietly belongs to someone else.

The uncomfortable question: What do you actually own?

Over the past two years, disputes over training data, open‑source code and AI outputs have moved into the mainstream of intellectual property practice. At the same time, frameworks such as the European Union AI Act are shaping global expectations about governance, even for small and medium enterprises that sell into, or source from, overseas markets.

Many Australian businesses now badge scoring engines, forecasting tools or assistants as “proprietary”, while relying heavily on open‑source components, permissive but conditional licences, and datasets whose legal status has never been tested. When a buyer’s lawyers start asking where the code and data came from, that label can peel off very quickly.

How “proprietary” tools quietly become derivative works

Modern AI stacks almost always include public libraries, pre-trained models and community datasets. That is efficient, but it matters which licences you have accepted. Some open-source terms include strong “copyleft” provisions that can force derivative works to share code, provide attribution or reveal source. If you have ignored those conditions, you may face licence claims or last-minute re-engineering when a deal is on the table.

Courts and regulators are also scrutinising how training data is sourced. Scraping content or ingesting confidential material without clear consent is increasingly cited in intellectual property disputes, and Australian parties using global tools are not insulated from that trend. For a buyer, a tool encumbered by conflicting licences and uncertain data provenance is not a strong asset; it is a risk that demands a discount or tighter warranties.

Due diligence: proof, not promises

In Australian transactions, technology and data diligence are now standard. Buyers and financiers expect at least:

  • A register of AI tools, models and datasets, with their licences clearly documented.
  • Evidence that attribution, confidentiality and client consent obligations have been honoured.
  • Confirmation that no “viral” open-source terms will compromise exclusivity in a sale.

If you cannot produce that evidence, your “proprietary” label will be treated as marketing rather than fact.

Shadow AI: value walking out the door

There is also the problem of “shadow AI”: staff using unapproved platforms to build automations and content. It looks like initiative, but it can mean your prompts, workflows and client information are being poured into external models where you have no exclusivity, and where your people can replicate the same value when they move on. That is not a moat; it is a leak.

Turning tools into real assets

The same disciplines that satisfy a sophisticated buyer also strengthen your position:

  • Audit every material AI tool and dataset, and confirm your legal basis for using each.
  • Clean up problematic open-source or open-data dependencies before they become a deal-breaker.
  • Document data provenance and staff use policies so that “shadow AI” is brought under control.
  • Capture evidence of how your tools improve margin, resilience or pricing power in ways competitors cannot easily imitate.

For Australian small and medium enterprises, the goal is simple: when you call an AI tool proprietary, you can prove it.For support in aligning your AI story with defensible intellectual property and International Valuation Standards, contact Kevin Lovewell directly on 1300 551 757.