Key takeaways 

• Directors will likely breach their directors duties if they rely solely upon a publicly available artificial intelligence (AI) tool to obtain advice about legal issues at board or operational level. 
   
• Australian corporate law is complex, and legal issues are often very nuanced. That is particularly so in disputes between shareholders and between directors where the catch-all legal principle of oppression is often overlooked.
   
• The difficulty for directors in using AI as a legal adviser is as much about not asking the right questions as it is about the accuracy and quality of legal output. 
   
• The Supreme Court of NSW has warned that:

[The directors] plainly directed questions to ChatGPT, which it had answered, which wrongly approached the question as one of employment law, rather than by reference to principles of oppression in Australian corporate law.

• Directors should seek legal advice where necessary.
   
• Executives also face liability risks when using AI for legal advice, as do government decision makers. The warnings for directors are just as applicable.
   
• There may be a wave of future litigation arising from directors’ and executives’ use of AI.

Directors and AI 

In our recent alert The unread board pack: can directors use AI to conquer information overload? 
we discussed the conundrum facing directors in the use of AI to digest their board packs and provide recommendations for decision making. 

The Chief Justice of NSW has since given a wide-ranging lecture about the use of AI by directors. He said it was important to recognise the reality that AI would become a tool for directors, but he echoed the words of caution identified by Justice Lee:

directors must be vigilant and disciplined in their use of AI… independence lies at the heart of good corporate governance, and great care must be taken by directors to ensure that the crucible of debate and productive conflict of ideas in the boardroom is not undermined or supplanted by AI.

A case recently determined by the Supreme Court of NSW provides an example of the risks of the use of AI by directors in their governance of their company.

Case study: Directors found liable for relying on ChatGPT for legal advice

Lanmar Pty Ltd is a company that provides services including engineering, asset management, advisory and consulting services to customers. A dispute between the directors resulted in a case heard in the Supreme Court of NSW and a recent judgment.

Ownership and control of Lanmar Pty Ltd was shared equally between three principals, Drew Landes, Martin Drebber and Peter O’Connor. They each (indirectly) held a third of the shares in Lanmar and were each directors.

As is common in small businesses, each of the three was also directly involved in the day-to day-business of Lanmar. However, as is also common in small businesses, the three principals fell into dispute about their various contributions and entitlements regarding the business. 

In this case, Mr Drebber and Mr O’Connor (majority directors) became unhappy with Mr Landes’ performance and contribution in relation to a bid for a substantial government contract. They thought Mr Landes “was a complete dud” and “he needs to go”. ^[1]

In this regard, the judge commented that:

It might be thought that, when the majority directors and shareholders of a company, in the midst of tendering for a substantial government contract, formed the view that a minority director and a minority shareholder “needs to go”, they would prudently take advice as to how to implement that result in a manner that minimised resulting damage to Lanmar. [73]

However, instead of prudently taking legal advice as to how to implement the plan to remove Mr Landes (including whether it was legally permissible at all), Mr Drebber and Mr O’Connor instead turned to ChatGPT. 

ChatGPT prepared a plan titled “HR problem resolution”. It created a “START framework” (comprising "situation, task, action, result, target") to set out the steps to remove Mr Landes.

There was a meeting between all three directors at which a document addressing the START framework was displayed. Mr Landes was advised by the other directors based upon the ChatGPT START framework that:

• They had serious concerns about Mr Landes’ performance and had lost confidence in him and his “input value was low”.
   
• They had “done their research” and knew they could “get rid of” Mr Landes.
   
• They had put in place a performance management plan which they expected Mr Landes to comply with and demonstrate significant improvement within 6 months.
   
• The responsibilities between the directors would be rearranged to exclude Mr Landes from any operational involvement in a significant tender.
   
• If there was no turnaround, the majority directors would consider altering the Lanmar directorship structure and introducing performance-based bonuses.
   
• Performance improvement would be monitored and assessed in monthly performance meetings.

For all intents and purposes, the meeting was in the nature of a performance management interview with a poorly performing middle-ranking employee who was in line to be terminated. There was no recognition or reference to the fact that Mr Landes was not an employee, but was director and shareholder. The legal issue was one of corporate governance, oppressive conduct and directors duties, not performance management and termination of an employee.

Oppressive conduct of company’s affairs

A particular and common issue in relation to disputes between shareholders in small companies is the general legal concept of oppression (by the majority on the minority). Broadly based upon common notions of fairness, the relevant test under s. 232 of the Corporations Act 2001 is whether the conduct is oppressive to, unfairly prejudicial to, or unfairly discriminatory against, a shareholder in that capacity or another. 

If such conduct is found by the court, the court then has very wide powers to make orders about how to remedy the oppression. This can include payment of compensation, making one or more parties buy the other one out at market value, the winding up of the company or the sale of the company to a third party.

Exclusion of a director/shareholder from management and operations in a closely held company (without a reasonable offer of exit) is a classic case in which oppression is often found to exist. 

Judge’s findings of oppression

In the case of the actions of the majority directors in Lanmar in seeking to get rid of Mr Landes, the judge had little trouble concluding that this constituted conduct of the affairs of the company (by the majority shareholders based upon the advice of ChatGPT) in a manner that was unfairly prejudicial, or oppressive, to the other director in his capacity as a shareholder.

The judge also found that the majority directors breached their directors duties by using ChatGPT rather than taking legal advice from a corporate lawyer in relation to Mr Landes’ position.

As a result, the judge ordered that all of the shares in the company be sold to one or more third parties. This resulted in all of the directors/shareholders of the company being forced out and the company facing a fire sale without its key human resources.

Criticism of directors use of ChatGPT

The judge found that the calamitous outcome was largely caused by the majority directors using ChatGPT to prepare and execute a legal strategy to “get rid” of the other director. In that regard, the judge found that:

• ChatGPT did not prove to be a prudent choice of adviser so far as matters of Australian corporate law were concerned.
   
• The directors appear to have wrongly framed the question to ChatGPT as one it took as an employment issue. This distracted attention from the risk of oppression in the course of conduct that was recommended by ChatGPT. In particular, the step of “forcing/commanding” plainly did not reflect the relationship between directors or shareholders under Australian corporate law.
   
• Not only did ChatGPT overlook the law relating to oppression, it also provided advice by reference to a clause in the company’s constitution that did not exist, and suggested that the directors divert business away from the company, something that is clearly a breach of company law. 
   
• ChatGPT did not advise the directors of the risk that the conduct it was recommending could result in the company being wound up for oppression and its value as a going concern destroyed.
   
• The directors should have taken legal advice from a lawyer as to the steps which were necessary to protect Lanmar’s interests against the course of conduct which they proposed, or as to the risks to Lanmar from that course of conduct. 

Lessons learned

Directors cannot rely upon publicly available AI platforms to obtain legal advice on how to deal with matters of corporate governance such as how to deal with a co-director and shareholder. Directors risk personal liability and put the company at risk if they do so.

Legal advice on corporate governance matters should be obtained from an appropriately qualified and experienced corporate lawyer.

The risks in using AI for legal advice and strategy is not limited to directors in respect of corporate governance. Any executive or government decision maker who relies upon AI for legal advice puts their organisation at risk.