The Federal Court’s AI Practice Note: Sensible Guardrails — or a Missed Opportunity?

The Federal Court of Australia’s newly issued AI Practice Note
(https://www.fedcourt.gov.au/law-and-practice/practice-documents/practice-notes/gpn-ai) marks a significant moment in the judiciary’s engagement with generative AI. It is, on its face, a pragmatic and timely response to the rapid integration of AI tools into legal practice.

But while the Practice Note is commendably cautious, it also raises a deeper question: is the Court merely reacting to risk, or shaping the future of legal practice?

A Necessary Intervention — but Fundamentally Conservative

At its core, the Practice Note does two things well.

First, it recognises the inevitability of AI in litigation. The Court expressly acknowledges that AI may improve efficiency, reduce costs, and enhance access to justice. That is an important and welcome judicial endorsement.

Secondly, it reinforces existing professional obligations in an AI context. The message is clear: AI does not dilute duties owed to the Court. If anything, it heightens them. The prohibition on misleading the Court is reaffirmed with particular force in response to AI “hallucinations”.

However, beyond these points, the Practice Note is strikingly conservative. It does not meaningfully reframe legal practice in light of AI — it simply overlays traditional duties onto new tools.

This is safe. But it is also limited.

The Disclosure Requirement: Transparency or Tactical Burden?

One of the more notable features is the Court’s ability to require disclosure of AI use — including what tool was used, how, and for what purpose.

While framed as a transparency measure, this raises practical and conceptual concerns.

From a practical perspective, the line between “use” and “non-use” is increasingly artificial. If a lawyer uses AI to refine structure, test arguments, or summarise authorities, does that require disclosure? The Practice Note provides examples, but not a coherent threshold.

From a conceptual perspective, the requirement risks turning AI into something exceptional — when, in reality, it is fast becoming as embedded as legal research databases or spellcheck.

There is a real question whether disclosure of AI use will remain meaningful, or devolve into a box-ticking exercise with little forensic value.

Confidentiality: Sensible Warning, Limited Guidance

The Practice Note places strong emphasis on the risks of inputting confidential or privileged information into AI systems, particularly public models.

This is undoubtedly correct — and necessary.

However, the guidance stops short of providing practical solutions. It warns against misuse, but does not engage with how practitioners are increasingly using enterprise or “closed” AI environments, nor does it articulate what level of technical assurance is sufficient.

In other words, the Court identifies the problem but leaves the operational response largely to practitioners.

For a profession already navigating uncertainty in AI governance, this may not go far enough.

The Unspoken Tension: Efficiency vs Authenticity

Perhaps the most interesting — and unresolved — tension lies beneath the surface.

The Court endorses AI for efficiency and cost reduction, while simultaneously insisting on strict human verification of all outputs. Practitioners must ensure that authorities exist, facts are provable, and evidence is admissible.

This effectively means that AI cannot be relied upon — only assisted by.

That position is defensible. But it also raises a difficult question: if the human verification burden approaches the cost of doing the work manually, where is the efficiency gain?

The Practice Note does not confront this tension. Yet it will become increasingly central as AI tools improve and expectations shift.

A Missed Opportunity for Judicial Leadership?

Ultimately, the Practice Note is cautious, principled, and necessary.

But it is not transformative.

It does not attempt to define best practices, establish safe harbours, or articulate a forward-looking vision of AI-assisted litigation. Nor does it meaningfully differentiate between levels of AI use — from administrative assistance to substantive legal reasoning.

In that sense, the Court has taken a defensive rather than leadership position.

Given the Federal Court’s influence, there was an opportunity here to shape national standards and provide clearer direction to the profession. Instead, the Practice Note largely defers that responsibility back to practitioners.

Where to From Here?

The Practice Note should be seen as a starting point, not an endpoint.

As AI becomes embedded in legal workflows, courts will inevitably need to move beyond general principles and engage with more difficult questions:

• What constitutes acceptable reliance on AI-generated reasoning?
• When does AI involvement affect the weight of evidence or submissions?
• Should there be differentiated standards for “closed” vs “public” AI systems?
• And ultimately — how does the concept of legal craftsmanship evolve in an AI-assisted world?

For now, the Federal Court has drawn a line: use AI if you wish, but you remain fully accountable.

That is a safe position.

Whether it is a sustainable one is another question entirely.

At SimplifyLaw, we are actively advising businesses and professionals on the legal, regulatory and practical risks arising from the use of AI — including confidentiality, compliance and governance issues. If you are navigating these challenges, feel free to reach out for a discussion.