“Replicants are like any other machine,” says Harrison Ford’s Rick Deckard in the 1982 film Blade Runner. “They’re either a benefit or a hazard. If they’re a benefit, it’s not my problem.” And while comparisons to Ridley Scott’s sci-fi thriller about android slaves throwing off their shackles are rife in the age of AI, the quote is no less applicable to the current moment in markets.  
 
On Monday, the Future Fund drove that point home with the release of its latest position paper, Portfolio Resilience: AI. Australia’s sovereign wealth fund has been a prophetic thought-leader on a range of emergent risks for markets and investors, and Portfolio Resilience: AI, with its talk of geostrategic competition and AI harnessed for military purposes, is useful for anybody trying to imagine the possible futures of our increasingly wild present.  
 
But it is the paper’s exploration of the more mundane AI threats of the very near future – say, next Wednesday, AD – that investors should spend their limited attention budget on.  
 
“As machine learning models replicate across asset management, market-making, and execution platforms, they increasingly train on similar datasets and respond to comparable macro triggers,” the paper says. “This creates conditions where a single adverse signal could trigger synchronised deleveraging across multiple institutions before human intervention is possible.” 
 
The risk described here is reminiscent of other instances of extremely funny, extremely value-destructive computer trading-driven sell-offs where a whole lot of people who thought they were smart suddenly discovered that they actually occupied the middle of the bell curve (as it pertains to the creation of esoteric computer trading models); think Long-Term Capital Management, or portfolio insurance.  
 
What to do about it? Assume the person down the street bought the same AI model as you, reads the same tech evangelical Substack as you and has the same big dreams about using machines to make money – or else get crushed in the rush to the fire exit.  
 
But that forward-thinking approach will only take you so far when it comes to another AI risk that keeps Colonial First State chief investment officer Jonathan Armitage up at night.  

His point (heavily paraphrased) is that the capital connections big AI and software companies have with each other, sealed with a handshake in an infrared sauna or luxury blood transfusion clinic, are increasingly arcane and create the kind of financial dependencies that would make even the most profligate keiretsu chairman hesitate; think Microsoft investing $13 billion in OpenAI only for OpenAI to spend most of that money on Azure, or Nvidia investing $100 billion in OpenAI to fund buildout of data centres that will be filled with Nvidia chips – etc.  
 
“And that ecosystem has yet to be tested,” Armitage told the Investment Magazine CIO Series podcast last week. “If one component of the system sees significant pricing risk, that is a bit of an unknown – and that’s coupled with the fact that business models are evolving very rapidly, but so are their capital structures…. That’s the sort of thing we’re spending a lot of time thinking about, and sometimes that’s at three o’clock in the morning.” 
 
The problem is that investors can’t just not own the stuff, given they’ll be flogged by their peers, and there are only so many apologetic annual letters that clients and members can read before they take their money elsewhere.  
 
The first step to dealing with it is to figure out exactly what you are exposed to, and where and how the portfolio can break; Portfolio Resilience: AI provides some handy examples of how the Future Fund thinks about what it calls access points, identifying four AI layers – hardware, infrastructure, platforms and applications – and the asset classes through which it owns them.  
 
Exercises like this are a handy first step when it comes to investigating the portfolio cross-exposures that many big asset owners are now grappling with; the next step is the gumshoe detective work of tracing counterparty relationships across equity, fixed income and private market books. Still, like any culprit in a detective novel, it’s likely that many will only reveal themselves under extreme stress.  
 
None of this is an argument for sitting the theme out, which is a luxury available to almost nobody managing money against a peer group or a performance test. But it’s a reminder that investors should care as much about the hazards of AI as they do the benefits – especially when Rick Deckard isn’t going to show up if the whole thing goes haywire.