In the past year social media panics caused “flash crashes” that bankrupted Silicon Valley Bank (SVB) and made giant banks such as Deutsche Bank wobble. SVB’s officers had unwisely purchased long-term U.S. Treasury bonds, which lost value in 2023 when interest rates increased dramatically. The bank failed when customers withdrew $42 billion in a single day after prominent venture capitalists, worried about its financial stability, used Twitter and other social media channels to encourage companies in their portfolios to withdraw their funds. The bank collapsed the next day when regulators took possession of it.

When such crashes happen, state regulators, as well as those at the Federal Reserve and the Federal Deposit Insurance Corporation (FDIC), are often caught flat-footed because they are not continuously monitoring banks for these kinds of faulty investments that make them unstable. As a result, regulators not only have to take over the already failing bank but other banks in the same ecosystem that made that same investment and are thus seeing the same losses. Regulators do this to stem subsequent runs that start as the customers of those related banks realize their bank is in the same boat as the one that’s failed and start pulling their deposits. Bank defaults are damaging to small businesses, including start-ups, as well as to ordinary depositors that depend on bank loans for the most basic financial operations, such as paying employees or suppliers.

The speed at which social media can spur massive financial movements is astounding. Before Twitter and Facebook, a spooked investor or customer would have to call, personally visit or even e-mail and text colleagues to urge them to withdraw funds from a troubled bank. This communication took days or even weeks, and while bank runs still happened, they were not always so catastrophic. Nowadays sophisticated clients can act as soon as they read a Tweet. Social media alerts everyone all at once, and a few clicks on a computer screen can wipe an account clean.

We have entered a new, dangerous era for finance where the centuries-old banking practice of investing in long-term assets backed by short-term deposits is failing. Our money is becoming more liquid, and our banking systems in the U.S. are unprepared for how online information—and misinformation—influences human behavior. The time has come to create banks that are more immune to bank runs and to promote the type of investing in which deposits cannot be quickly withdrawn.

In addition, the Federal Reserve and related agencies must start to continuously monitor banks to catch destabilizing actions such as those that happened at SVB before they turn into bank runs. Instead of protecting the failing bank, agencies should protect the economy and smaller-scale consumers. Lending is risky by nature, and there will always be issues, but surprisingly, both U.S. bankers and regulators know that today’s complex regulations could not have stopped what happened to SVB. Regulators and banks even expect repeated failures yet only have emergency procedures in place to respond. It is as if you know flu season is coming every year, can’t get a vaccine and end up in the emergency room every time, no matter how sick you get.

Today’s banking system is a chaotic patchwork of old and new financial instruments, legal structures, computational advances, all-too-frequent crashes and new regulations that don’t address this problem. Because current banking rules do not (and cannot) protect against default, insurers expect that the average bank has a 1 to 2 percent chance of failing. These are what we refer to as credit default swap rates, and they are similar to fail rates for other stable companies in other industries. Still, such a rate is rather problematic for banks, given their systemic importance.

The first significant obstacle to creating banks that are more resistant to runs is that we mix narrow banking (the payment and deposit functions of banks) with fractional banking (loaning out depositors’ money to what are inevitably risky deals, such as mortgages, commercial loans or start-up financing and keeping only a fraction of the depositors’ money for day-to-day business). Losses are usually invisible to depositors because a sufficient fraction of their money is still available to keep daily business going, and profits from repaid loans fill in the holes caused by bad debt.

The second barrier to reliable banking is that today’s banks will always fail when enough depositors decide to withdraw their money at once because their bank cannot convert their long-term assets into cash fast enough. Social media’s power to get large numbers of people to act simultaneously means that bank runs happen too quickly to stop. This risk gets multiplied by the fact that banks have overlapping financial obligations and loan portfolios so that the failure of one bank can propagate, and banks can begin falling like dominoes.

The first part of a cure is to break commercial banks into narrow and fractional banks, to separate day-to-day financial services from the business of making loans. This is similar to how lawmakers behind the Glass-Steagall Act of 1933 spread risk by breaking banks into commercial and investment entities after the financial collapse that led to the Great Depression. Narrow banks that do only day-to-day operations and make their money from small fees can’t go broke because of runs on another bank. If the narrow bank’s operation and liabilities are legally distinct in segregated accounts, fractional bank clients will not have a claim on the funds of narrow banks’ clients. If the loan bank fails, day-to-day banking operations can continue, protecting the average consumer.

A second part of a cure is illustrated by how investment firms, such as the private equity firm Blackstone, protect themselves from runs on their investments—they limit the amount of money large depositors can withdraw at any one time. These arrangements are legally binding; clients agree to them willingly because Blackstone promises very good returns, so investors forfeit some flexibility in return. These limitations make Blackstone relatively immune to the panics and fads of social media.

If banks have some breathing room, they can sell enough illiquid assets to repay the depositors. Similarly, in 2008 most of the companies that the federal government bailed out didn’t fail, and that’s because they had breathing room—they could sell enough assets to eventually repay their emergency loans.

The final part is for regulators to start digitally monitoring banks wholly and continuously instead of the current system of annual audits and stress tests. After all, banks settle their books at least once a day, so auditors could easily flag problems in real time. Contemporaneous audits of bank books will improve banking behavior and discourage some riskier activities these firms might otherwise engage in. Continuous auditing also minimizes the need to seize the bank, destroy much of the value of the bank’s investments and scare everyone unnecessarily.

For instance, if the Federal Reserve had been continuously auditing SVB, it would have seen much earlier than the public that the bank’s investments had gone south and would have been able to give SVB time to turn enough of its investments into cash to shore up its basic services. Of course, this breathing room would come at a price: the bank would pay an appropriate penalty, which would keep it motivated to fix its bad debt situation, thus avoiding the bank run and liquidation.

As nations adopt digital money, artificial intelligence advisers start managing our finances and instant payment services, such as the recently launched FedNow, become more common, the need for banking overhaul becomes even more urgent. We are essentially able to move money faster than banks can. The more liquid our money becomes, the more critical it is to retool our banking systems to prepare for social media panics and the other ways that our digital lives influence money.

This is an opinion and analysis article, and the views expressed by the author or authors are not necessarily those of Scientific American.