Investor Sues After an AI’s Automated Trades Cost Him $20 Million
BY Kristin Houser
The first-of-its-kind case could shape the future of AI legislation.
Back in 2016, Futurism noted that artificial intelligence was creeping into the world of finance, analyzing market data in real-time to autonomously execute trades for investors.
Now, Hong Kong real estate tycoon Samathur Li Kin-kan is suing a company that used one of those trade-executing AIs to manage his account, causing millions of dollars in losses — a first-of-its-kind court case that could help determine who should be held responsible when an AI screws up.
According to a Bloomberg story about the case, Li met Raffaele Costa, CEO and founder of Tyndaris Investments, in March 2017, at which point Costa told Li his company was launching a robot hedge fund controlled by a supercomputer named K1.
Li expressed interest in the fund, so Costa started sharing simulations with Li that showed how K1 could make double-digit returns on investments.
Li was apparently impressed, because he agreed to let K1 manage $2.5 billion, with the goal of eventually increasing that to $5 billion.
K1 began managing Li’s investment in late 2017, and by the time February rolled around, the AI was losing money regularly — on one particularly bad day, its decisions cost Li more than $20 million.
Li pulled his money from the account before the end of that month. Then he filed a $23 million lawsuit against Tyndaris, alleging that Costa exaggerated K1’s abilities.
Tyndaris’ lawyers, meanwhile, deny that allegation, asserting that the company never made any guarantees that the AI would make money for Li. They’re now suing him for $3 million in unpaid fees.
According to Bloomberg, these lawsuits are the first known examples of humans clashing in court over financial losses causing by an AI-powered trading system. However, it likely won’t be the last time we see the judicial system forced to tread new territory thanks to the emergence of AI.
Should it be the person who wrote the code? The person marketing the AI? Or are end users responsible for the outcome given that any new technology is certain to be a bit wonky at first?
The outcome of the K1 trial, which is scheduled for April 2020, might finally start to set some legal precedent.