When Silicon Valley Bank teetered on the brink of collapse in March 2023, all eyes turned to social media. The California-based bank was a favorite of the tech industry, including venture capitalists and startup founders who began to raise concerns about the firm’s balance sheet on platforms like Twitter, Discord, and Telegram. Panicked customers withdrew $42 billion on a single day, which led to the second-largest bank failure in U.S. history.
Just as the 2008 crisis rewrote the playbook on bank failures, regulators are still working to parse the circumstances that led to the undoing of Silicon Valley Bank and two other institutions that collapsed the same month: The crypto-focused Silvergate and Signature. Academics and politicians began to immediately finger social media as a potential culprit, with House Financial Services Chair Patrick McHenry (R-N.C.) describing Silicon Valley Bank as the “first Twitter-fueled bank run” in a statement on March 12, 2023.
Despite such rhetoric, social media was likely a symptom more than the cause of Silicon Valley Bank’s failure, which came about due to poor asset management amid rising interest rates. Nonetheless, government agencies have taken notice of the panicked tweets that preceded the bank’s collapse and are responding to the rapid-fire spread of information with new strategies. These include one U.S. banking regulator working with a blockchain intelligence platform called Inca Digital to monitor social media for discussion about its supervised banks, starting in late 2023.
As banking officials move toward a world of nearly 24/7 operating hours, monitoring social media is expected to become a vital tool for predicting the next potential failure—even if platforms like Twitter are not the main catalyst. “Whether you think social media is causing it or not, it might help [regulators] predict it faster,” said Julie Hill, a banking expert and the incoming dean of the University of Wyoming College of Law. “You might be able to identify that a run is happening faster and give yourself more time to address it.”
Blockchain meets TradFi
A lawyer by training and veteran of the U.S. Air Force, Inca Digital CEO Adam Zarazinski founded his firm in 2018 with the goal of helping companies cut through the noise of the blockchain space. One of Inca’s most popular tools was its social media monitoring, which companies used to track platforms like Twitter to get advanced notice of the latest hack or depeg. “All of crypto lives on social media,” he told Fortune.
Believing that trends in crypto presage those in traditional finance markets—the rise of meme trading is one example—Zarazinski sought to apply this logic to social media. His thesis paid off when, in a conversation with one of the banking regulators, they asked if Inca’s social media tool could be reworked to monitor chatter about banks, rather than blockchain protocols. He declined to share which regulator due to a confidentiality agreement, although the contract began in late 2023.
With a staff of around 40, Inca has been adapting its software to track social media chatter surrounding banks. Rather than sentiment analysis, which uses natural language processing to gauge the tone of text—such as whether it’s intended to be sarcastic or serious—Inca uses AI to identify topic models, or what people are talking about based on the context of posts. Inca trains the model to identify when discussion around a specific topic reaches a threshold and then notifies the client based on different metrics, such as the number of posts about the risks of a specific bank over a certain period.
As a retroactive test, Inca applied its crypto Twitter model to Silicon Valley Bank, which Zarazinski shared with Fortune. The model showed a bar chart breaking down posts from the time into two categories: risk-related and not-risk-related, beginning on Monday morning. Posts begin to tick up on Tuesday morning, and then peak for the first time around 4:00 pm on Wednesday, when Silicon Valley Bank began to issue a series of announcements about its portfolio. According to Zarazinski, that’s when Inca would’ve alerted the regulator. Withdrawals began in earnest the following day.
‘Hard to untangle’
Experts are still split over the role that social media played in the run on Silicon Valley Bank. A widely shared paper by five academics in the wake of the collapse found that social media amplified bank run risks, with the intensity of Twitter conversation predicting the losses in a bank’s share price, which represented broader sentiment about the bank’s stability. “Given the increasingly pervasive nature of social communication on and of Twitter, we do not expect this risk to go away,” they wrote.
Hill, the University of Wyoming professor, said that academics still don’t have a well-accepted theory for what leads to bank runs in the first place, let alone the impact of social media. “If we were really good at that, we would just eliminate them altogether,” she joked.
The narrative that Silicon Valley Bank was the first “social media” bank run spread after its failure, but Hill remains unconvinced, especially given the reality that the tight-knit community of depositors at the bank were likely communicating mainly through private channels like text messages. “I’m not saying that banks or regulators shouldn’t be worried about what’s going on on Twitter, but I just think it’s very hard to untangle,” she told Fortune, adding that social media still represented a new tool for observation.
Jess Cheng, a former senior counsel at the Federal Reserve Board and partner at the law firm Wilson Sonsini, said that the recent crises have challenged many of the first principles of bank supervision, with social media adding a new wrinkle. “No one was thinking about who was tweeting what,” she told Fortune.
As the Federal Reserve explores opening up its settlement service to operate 22 hours per day, seven days per week, the established playbook could grow even more difficult, as regulators may no longer have the weekend to clean up failed banks, as they did with Silicon Valley Bank. The constant beat of social media, and its viral spread, would only fan the flames.
Cheng said that unlike the 2008 crisis, which was largely caused by structural failures at banks, the recent crisis was mostly driven by a flurry of rapid-fire customer withdrawals—which occurred faster than ever thanks to the ubiquity of mobile banking. While Twitter chatter would not have changed the reality of collateralized debt obligations in 2008, it may have caused a flood of VCs to panic. “A banking agency can’t control what people make of what they see on Twitter,” she told Fortune.
Inca is not the first time a banking regulator has dabbled in social media. In January, Reuters reported that the European Central Bank asked some banks to monitor activity on social media. Inca’s approach seems to go further, however, with the firm tracking several platforms, including Twitter, Discord, Reddit, and Telegram.
Since starting the contract, Zarazinski said that no alerts have yet been triggered, although Inca’s model only recently began to monitor smaller and mid-sized banks. One recent candidate was New York Community Bank this March, although Zarazinski said that concerns over its stability were already well-known. But Inca will be ready for the next one.
“Silicon Valley Bank was a real slap in the face for everybody to wake up to how information spreads through social media,” he told Fortune.
