• Published: Dec 12, 2023
  • share on linkedin
  • share article

It’s been a busy month for public failures of governance. Let’s start with possibly the most interesting, OpenAI. We could usefully begin with a reflection on the nature of irony, which apparently works like a hyperbolic function. The more ironic a situation is, the less likely those involved are to realise it.

Someone had the wit to ask ChatGPT for an opinion on the governance structure at OpenAI and this was the answer: 

“The corporate structure as outlined shows a complex relationship between different entities which could pose several risks. 

  1. Conflicts of interest. The presence of a shared board of directors across not-for-profit and for-profit entities can lead to conflicts of interest, particularly when decisions might benefit one entity over the other. 

  2. Regulatory scrutiny. The structure shows a mix of nonprofit and for profit entities which could attract regulatory scrutiny, especially if the flow of funds between these entities is not transparent and compliant with tax laws. 

  3. Complexity in governance. The structure appears complex with multiple layers of control and ownership, which can lead to difficulties in governance and decision making, potentially slowing down operations.

  4. Reputational risks. If one part of the structure is involved in controversial activities, it could have a spillover effect on the reputation of the other entities due to their interconnected nature. 

  5. Legal risks. The structure may face legal challenge if the separation between the nonprofit and for-profit activities is not clearly maintained.

  6. Mission drift. There’s a risk that the for-profit motivations could overshadow the non-profits’ mission, leading to drift from original purpose. 

These are broad risks, and the actual risk level would depend on the specifics of how this structure is managed and regulated.” 

Of the many justifiable fears associated with AI, the danger of sensible advice is usually not on the list. 

From a governance perspective, there are several observations. What started as an optimistic, altruistic nonprofit venture is now a valuable company estimated at $US86b. The original governance structure was not designed to represent shareholders but designed with safety in mind. Responsible development of the software was the intent—or, as some have said, to prevent an AI apocalypse. The board was three OpenAI executives and three people who may have been the right fit early on but not perhaps for something that has ballooned in value and commercial attraction.   

The company is now 49% owned by Microsoft, which has invested $US10b. Yet, most unusually, Microsoft had no seats at the table. Nor, it seems, was Microsoft or any other investor consulted ahead of the decision to fire CEO Sam Altman. That is barely believable. 

The real reason for the firing is yet to be revealed but media commentary would indicate division around the speed of development and consequent safety. Some have referenced a nascent new product called Q or Q Star—a radical leap forward in the technology. Perhaps this is where the CEO was failing to be “consistently candid in his communications” with the board. 

Herein lies another problem. American boards favour executive directors and chair/CEO combinations. Here Mr Altman seems to have forgotten he was a member of the board and hence had a range of specific duties to the organisation as a whole. There is a strong possibility that he was a negligent director. This point seems to have been missed.  

In this case, the board chair was Greg Brockman, previously described as president, chairman and co-founder. Apparently, he was not even on the video call that made the firing decision. Mr Brockman seemingly decided that the multiple inherent conflicts he faced in this situation were best dealt with by excusing himself from the process—a structural failure at a minimum. 

Microsoft have moved to shuffle seats at the table, but the underlying structural problems will not go away. Quite what Microsoft intended with its approach also remains unclear. It was prepared to hire Altman, Brockman and hundreds of others who wished to leave OpenAI thereby essentially destroying a company it owned nearly half of. 

It is crucial that this company and others like it have effective governance. The battle here is noted as between effective altruists and effective accelerationists—those working towards what is known as artificial general intelligence, something as versatile and adaptable as the human mind.  

Oversight of this cannot be placed within weak governance structures and inadequate regulation. The intent behind the original structure is laudable but it is under significant pressure and needs serious review. As Altman himself has said, “the worst outcome from the adoption of artificial general intelligence could be lights out for all of us”. [2] 

Rugby (again) 

Hamish McLennan finally lost his battle to retain the chair’s role within Australian Rugby. The cynic might say that following the World Cup debacle, someone had to go. Some of the changes he was pushing for such as centralisation of the high-performance function and some commercial matters we already see in New Zealand.  

But Rugby Australia is a federal system, and the state unions have the votes. As we have seen locally, these structures are very difficult to lead. One suspects the problem was more about style and role confusion. A chair is not the leader of the organisation nor the boss of the board. McLennan seemed to be behaving as an Executive Chair and his pronoun use was weighted to ‘me’ and ‘I’, rather than ‘we’ and ‘us’.  

He was accused of “acting outside his role as a director, exerting an undue influence on the operations and executives of Rugby Australia”. [3] Chairing large federal structures is complex and relies on personality and quiet persuasion rather than bluster. Wrong chap in the wrong seat seemingly. 

Australia again (OPTUS) 

Another ‘disaster, so someone had to go’. A national outage crippled sections of the country’s key infrastructure. Optus was criticised for poor communication and a lacklustre response during the event. Seemingly there was no disaster management plan for this kind of event. Surely the board should have expected such a plan to exist.  

But there is no real board for Optus Australia. It is a subsidiary of Singtel Ltd in Singapore. Paul Sullivan is cited as chair and a couple of senior executives are noted as Managing Directors. The Singtel annual report [4] notes this entity as an advisory committee only, raising serious questions about the effective governance of an organisation whose services are critical to the functioning of the country. 

 



Notes

  1. David Moore. Evolutionary Logotherapist. https://www.quora.com/What-is-hyper-irony 
  2. The Guardian November 23  
  3. Foxsports November 19  
  4. Singtel annual report