FSB's AI Warning: Amplifies Vulnerabilities Despite its Benefits
Remember programmatic trading induced "flash crashes?"
Better late than never, the Financial Stability Board (FSB) finally acknowledges that AI may negatively impact stability and amplify vulnerabilities.
If you thought using AI to control nuclear armament was a bad idea, wait until you see the FSB's list of problems with AI.
The truth is that the FSB doesn’t know AI’s impact. AI is so new and unevenly disbursed within the financial system that no one, including the FSB, can determine which parts are the most vulnerable.
That isn’t to say that we should fear a crash tomorrow, but we should be mindful of AI's potential impact on stability.
Remember programmatic trading? This predicament will seem eerily similar for those old enough to remember flash crashes and the impact of programmed trading on markets.
Like AI, no one knew how widely disbursed programmed trading was or what would happen if all the programs sold stocks at once until it happened.
I suspect our next AI crash will be similar!
👉TAKEAWAYS
🔹 Third-party dependencies and service provider concentration:
Greater reliance on and market concentration among AI service providers can increase systemic third-party dependencies in the financial sector.
🔹 Market correlations:
AI-driven correlation vulnerabilities could interact negatively with increasing levels of automation in financial markets and greater speed and accessibility.
🔹 Cyber threat:
LLMs and GenAI could enhance cyber threat actors’ capabilities and increase the frequency and impact of cyber attacks. Greater usage of specialized service providers exposes FIs to operational risk from cyber events affecting these vendors.
🔹 Model risk, data quality, and governance:
Wider uptake of complex AI approaches could increase model risk for FIs that cannot effectively validate, monitor, and, when necessary, correct AI models.
How will AI trading increase the overall share of trading done by machines and will correlations increase as a result of machines acting in concert? Above from IMF
👊STRAIGHT TALK👊
I am the first to grant that your broker’s chatbot won’t likely cause a crash and that many AI uses are benign.
That said, many new AIs are quietly being brought online for their predictive power, and we don’t know where they are concentrated. We do know that most AI’s are black boxes, and their users can’t fully predict their actions even more so than we saw with programmatic trading.
For one example, look at Ant’s latest product for corporate treasury management, which claims to predict currency exchange rates on an hourly basis to help reduce transaction costs. HERE
While I don’t see treasury transfers tanking the market anytime soon, they’re a great example of how AI and pricing are connected and likely already more deeply buried in our financial system than most of us know.
AI is great, but someone better get busy regulating it before we repeat another flash crash or worse.
Readers like you make my work possible! Subscribing is free, and I use the same business model as public broadcasting, where you can get all of my writing for free. If you like the content, please buy me a coffee by subscribing. Thank you!
Sponsor Cashless and reach a targeted audience of over 50,000 fintech and CBDC aficionados who would love to know more about what you do!
