Key points

Recent Australian commentary and case law signal where director liability for AI use is heading, and New Zealand directors should take note. While these materials are not binding here, the New Zealand Companies Act creates similar exposure. The core message is simple: directors cannot outsource their judgement to AI and still claim the legal protections that depend on them exercising it.

• Directors cannot delegate their powers directly to AI.
• Boards must be able to justify their trust in delegates who use AI, including that delegates are using it responsibly.
• Directors should not blindly rely on AI-generated opinions or recommendations.
• AI notetaking tools in Board meetings carry legal risk and could chill frank discussion.
• Boards that shun AI altogether may face greater risk; without sanctioned use, “shadow” AI fills the vacuum.
• Boards should adopt or refresh internal AI policies covering disclosure, verification, and boardroom use.

A Senior Judge’s warning

The Chief Justice of New South Wales, Hon. Andrew Bell AC, has warned that directors who outsource their thinking to AI tools risk losing certain legal protections.^[1]

Speaking extra-judicially in a recent Law School lecture, Bell CJ warned that where AI has played a prominent role in the making of Board decisions, statutory safe harbour provisions may not be available to directors. He cautioned that AI can resemble fraudsters, noting automated-decision making failures like the robo-debt scandal in Australia and the Post Office Horizon IT fiasco in the United Kingdom as cautionary examples.

These comments focus on Australian law and are not binding in New Zealand. However, the New Zealand Companies Act creates similar exposure, meaning the reasoning is directly relevant here. Bell CJ’s comments also follow the Australian Federal Court’s judgment in ASIC v Bekier, the first Australasian decision to address directors’ use of AI.^[2] In Bekier, the Court said that any AI use by directors should be controlled and transparent.

Delegation of directors’ powers

Can directors delegate to AI?

Both the New Zealand and Australian companies’ legislation allow directors to delegate their powers to other people.^[3] Because AI does not have legal personhood, it cannot be a delegate. Directors’ powers cannot be assigned directly to an AI tool.

What about delegates who use AI?

But what if a human delegate uses AI to perform or assist with a delegated task? Section 130(2) of the Companies Act protects a Board from liability for a delegate’s exercise of power, provided the Board believed on reasonable grounds that the delegate would exercise the power consistently with the directors’ duties in the Act, and has monitored the exercise of that power by reasonable methods properly used.^[4]

Bell CJ’s view is that Boards will need to justify why their trust in a delegate relying on AI was reasonable, considering AI’s limitations, including its “scope for seductive self-confidence and sycophantic agreement, bias, or hallucination.” He said this analysis will depend on the nature of the decision and the type and quality of AI used, including whether the AI was a closed and curated database or if it was an “open” system.

Bell CJ suggested that to help retain the safe harbour protections, Boards should require management to disclose whether AI was used in preparing Board reports, how it was used, what verification was undertaken, and whether the information is considered reliable.

Reliance on AI generated information provided by others

Boards frequently rely on information prepared by management, professional advisers and experts. The Companies Act provides a safe harbour for directors who rely in good faith on such information, where the director believes on reasonable grounds the information prepared is within the person’s competence.^[5]

Bell CJ focussed on what happens when those advisers use AI to help prepare their work. He concluded that while management and professional advisers are entitled to use AI to help prepare their work, the belief by the director as to the person’s competence may extend to competence in the responsible use of AI. Directors may also need to make proper enquiry where it is indicated that the person who has used the information has relied on AI.

Business judgement rule

Although New Zealand has no statutory business judgement rule,^[6] the courts have recognised that they should be wary of second-guessing business judgements that are made by directors in good faith.^[7] Bell CJ considered it “extremely doubtful”, however, that a director who simply adopts the advice of an AI tool could shelter behind the business judgement rule. Ultimately, a director who outsources its thinking to AI is unlikely to be viewed as genuinely exercising his or her own judgement.

Board meetings

Bell CJ described the boardroom as the crucible of debate, dissent and challenge. He warned that AI may erode this dynamic by way of its characteristic “overconfidence” that tends to engender over-reliance and deference in its users. In particular, he expressed caution that AI notetaking tools could have a chilling effect on free and frank discussions. While these tools lend to convenience, they risk inhibiting robust debate, particularly given the discoverability of boardroom transcripts in legal proceedings.

Also see our previous article on AI notetaking in the boardroom: Simpson Grierson - Directors take note: the legal pitfalls of AI in the boardroom.

The “shadow” AI problem

The Judge did not, of course, counsel not using AI at all. Resisting it and the implementation of tailored AI platforms simply pushes its use underground, with directors and employees instead turning to personal accounts and public large language models, which carry serious cybersecurity and confidentiality risks.

Bell CJ’s concern about “shadow” use of AI also echoes the opinion expressed in Bekier. There, the Court cautioned that “boards should discuss and deliberately govern any AI use by formal adoption of policies, rather than just wink at informal “shadow” use.” There is, however, “… considerable potential for AI, if appropriately utilised, to assist directors in the discharge of their duties.”

New Zealand’s regulatory landscape is still developing. The Law Commission’s review of directors’ duties is due to be reported in 2027, and we expect AI to feature in this work. 

If you would like advice around AI governance or directors’ duties, please get in touch with one of our experts.

Special thanks to Georgia Hughey for her assistance in writing this article.