Last week, the Twitter accounts of Joe Biden, Elon Musk, Jeff Bezos and other high-profile users were hacked in a bitcoin scam. The hack brings up several nightmare scenarios about how social media platform could be used to foster carnage and chaos.

What if the hackers had been seeking not profit, but conflict? What if, rather than tweeting a crude bitcoin hoax, they had tried to provoke war? What if the tweet came from a government account, such as U.S. Strategic Command? Our team at King’s College London has been researching precisely those scenarios. Here’s what you need to know about the risks of conflict escalating via Twitter.

The United States is asymmetrically vulnerable to escalation by tweet

A group of hackers apparently used an internal Twitter tool to take over the accounts of prominent individuals and, from their accounts, to tweet a bitcoin scam that raised over $100,000 in two hours. Twitter shut down hundreds of thousands of verified accounts to stop the scam from spreading further.

The United States is particularly susceptible to Twitter disinformation. About 20 percent of Americans use Twitter, while only 6 percent of Russians and 1 percent of Indians do. This means Americans are more likely than any other nationality to read tweets and will be more exposed to disinformation on the platform.

Twitter could contribute to crisis escalation in three ways

First, governments could inadvertently escalate a crisis because of miscommunication and misperception. Collateral messaging — a term meaning a message intended for a domestic audience but interpreted differently by an international one — is a long-standing problem in diplomatic communications, and particularly challenging over Twitter. For example, when President Trump referred to North Korean leader Kim Jong Un as “Little Rocket Man,” he probably intended to project a strong image to the U.S. public. But North Korea would have perceived it as baiting and aggressive.

Twitter could also thicken the fog of war and sow confusion. That’s complicated by the fact that leaders can generate information overload, making it difficult for the public, allies and adversaries to differentiate background noise from the intended message.

Over the 10 days of the U.S.-Iran crisis this year, Trump tweeted over 181 times, 69 of which were about Iran. Conversely, Iranian Foreign Minister Mohammad Javad Zarif, a relatively active tweeter, tweeted about the crisis just 12 times. Trump generated a large amount of background noise, making it difficult for outsiders to sift through the messages and understand his true meaning.

Next is intentional escalation. A malicious government might use Twitter to distribute disinformation or amplify alarmist stories to try to influence another government’s behavior. For example, in a NATO-Russia conflict Russia could tweet disinformation to provoke an American response, such as nationalist calls for military intervention based on fake news.

And, finally, a third party, such as a group of hackers, could push governments into a conflict through social media. This could be done through a leak of sensitive materials, disinformation or a politically motivated hack against a major political party. Such nongovernmental groups are increasingly active on social media. Bot activity is on the rise, and often observers can’t tell if they’re run by governments or outside groups.

A Twitter shutdown will not solve the problem

Several of Twitter’s decisions have a public impact. The company’s response to the recent hack was to freeze over 300,000 verified accounts. While necessary, such a blanket shutdown could be dangerous in a crisis.

People rely more on social media during times of uncertainty, research finds. In a survey of more than 25,000 consumers in 30 markets conducted from March 14 to 24, the consulting firm Kantar found that the coronavirus pandemic prompted a 40 percent increase in global use of Facebook and WhatsApp. One study found that during a major U.K. snowstorm in 2010, Twitter activity spiked as snowfall increased, and that government officials used the platform to respond directly to public queries. In another example, research found Twitter was essential in helping direct humanitarian aid during the 2010-2011 electoral crisis in Côte d’Ivoire.

Twitter’s internal security — or lack thereof — affects not just its profitability but the public. Last week’s hackers told Vice magazine that they had paid a Twitter employee for access to the accounts, saying, “We used a [Twitter] rep that literally done [sic] all the work for us.” This would be particularly troubling during a crisis.

Twitter could mitigate many of these risks by developing a more sophisticated response plan to future hacks, rather than a massive shutdown of verified accounts. It could also share information about the most recent hack to inform research into how to prevent future hacks and how to respond to the next one, particularly if it occurs during a crisis.

Governments need Twitter crisis management plans

Governments may wish to create a crisis management plan to mitigate the risks of Twitter escalation. Our research suggests precrisis management is particularly important, because once a crisis begins there may not be time to develop a plan in real time.

Toward that end, governments could develop an interagency guide of best practices for tweeting during crises and for coordinating social media activity across government accounts to avoid working at cross purposes.

This year’s U.S.-Iran crisis can help demonstrate how mixed messages could escalate a crisis. On Jan. 3, Secretary of State Mike Pompeo sent 23 tweets about meeting with regional allies and expressing the U.S. desire for a peaceful resolution to the crisis. The next day, Trump tweeted that the United States had targeted 52 Iranian cultural sites and threatened that Iran would “BE HIT VERY HARD AND VERY FAST.” Which message were Iranians to believe: Pompeo’s calls for peace or Trump’s threats?

But once a crisis starts, our research suggests that the most valuable step governments can take to avoid escalation is simple: Stop tweeting individually and only allow official pronouncements. This would ensure the U.S. government is sending coordinated and consistent messages to minimize the risks of inadvertent escalation. Otherwise, next time the damage could be much more serious than $118,000 in bitcoin.

Heather Williams (@heatherwilly) is a lecturer in the defense studies department and the Centre for Science and Security Studies (CSSS) at King’s College London.