Since the advent of social media, guardrails have been in place to deal with extremist content and helping prevent users from becoming radicalized. The removal of many of these safety mechanisms recently was based on arguments prioritizing freedom of speech over the damage done by excessive hate content.
Algorithms' aggressive promotion of content stoking the worst impulses of an ever-growing number of radicalized individuals online – including some with mental health issues – has overtaken the guardrails that remain.
What is currently happening online is no surprise to MEMRI – which, for almost two decades, has worked with social media companies and government agencies on understanding how online radicalization can lead to attacks. As I wrote for Forbes in March 2026, this algorithmic radicalization is something that tech companies must tackle.
In their manifesto, the 18-year-old perpetrators of the May 18 shooting at a San Diego mosque, who according to authorities were likely radicalized online, set out their enmity towards many races and religions and referenced known online extremist influencers. They are part of an entire generation impacted by the toxic social media content fed to them by algorithms.
Days later, on May 23, 21-year-old Nasire Best tried to enter the White House, shooting two bystanders outside; he was shot and killed by the Secret Service. A former classmate later said that Best, "light-hearted" and "energetic," had "never been an aggressive person." Best, who had been in financial difficulties and was evicted from his apartment, was active on Instagram and TikTok; in July 2025 he had posted online that he was the "real" Osama bin Laden and that he wished to harm President Trump.
The perpetrator of the April 25 attempt to assassinate President Trump at the White House Correspondents' Dinner had posted selfies, complete with weapons, just before his attack. He had sent a manifesto detailing his plans to family and a former employer. His online activity on Bluesky, YouTube, and X appeared to encourage violence against President Trump.
Other young would-be terrorists have openly acknowledged their online radicalization – including one of the New Jersey teens set to plead guilty after allegedly pledging allegiance to ISIS and plotting attacks, who is blaming social media algorithms for it. Then there is the Ohio 14-year-old targeted for recruitment by jihadist extremists online who was found to be planning an attack. Others too have been arrested this past year for making online threats to politicians and religious institutions.
These individuals seemed to have believed that social media had made them important, prompting them to post videos, manifestos, and other documentation to secure their place in history before carrying out their crimes.
The pattern is consistent: The perpetrators held a grievance, emanating from anger, perceived isolation and injustice, or personal failure. Their grievance was reinforced and justified in today's online environment, smoothing the path to choosing an actual target.
Seeking attention and becoming more radicalized in the process, individuals like these strive for more views, more engagement, and more followers online. Often, they do not realize that observers pity them and family and friends are increasingly worried about their unhinged online behavior.
A growing body of research on the psychology of people getting lost online instead of living in the real world indicates that this behavior is similar to other addictive behaviors involving dopamine hits and a distorted sense of reality. Global meta‑analyses estimate that 14% of the global population meet the criteria for Internet addiction – U.S. rates tend to fall within similar ranges – and 5-10% of the U.S. population is at risk of serious social media addiction. This addiction frequently coexists with other mental health disorders.
As some scholars have pointed out – and as MEMRI has seen in its research and monitoring on this issue – this behavior eventually alters brain function and impacts impulse control – as reflected in the increasingly fanatical and violent content they are posting.
A recent Rand Corporation study, "Manipulating Minds – Security Implications of AI-Induced Psychosis" states that reports of AI-induced psychosis suggest that large language models (LLMs), and, in future, artificial general intelligence systems (AGI), might be capable of inducing or amplifying delusions or psychotic episodes in users. The study looks at whether AGI could be weaponized to induce psychosis at scale to create significant national security threats.
According to a recent study by the Charles F. Kettering Foundation and Gallup, Americans who spend at least five hours a day on social media are less accepting of those who do not share their beliefs, and more open to political violence. They may have been "more disaffected to start with," according to Jaime Settle, an associate professor of government at William & Mary. Derek Barker, senior program manager for research at the Kettering Foundation, speculated that "social media might be reinforcing the tendency to associate with like-minded people who are reinforcing these more extreme beliefs."
Warning signs of radicalization include rapid changes in the content such individuals are posting, particularly obsession with a specific ideology of hate, sudden hostility, justification of violence, or frequent attacks on people they know. Other red flags include abandonment of long-held beliefs and a clear ideological shift; fixations and indications of an obsessive desire for score-settling and vengeance for perceived wrongs; expressions of intense hatred of others, and hopes for harm to come to them or even expressions of intent to do this themselves. Such transformation, whether sudden or gradual, is the hallmark of online radicalization – and yes, people who post daily about their hatred should be considered a threat.
Many have observed the Internet's key role in the growing extremism – yet we seem to be sliding backwards into a less regulated and more dangerous online space. The recent X announcement that it would block UK access to accounts linked to terrorist groups banned in the country is a positive move, as was YouTube's shutting of Iranian accounts targeting Western youth. Another positive move was the introduction in May of the Jewish American Security Act by Senators Jacky Rosen and James Lankford; it aims to force large social media platforms to publicly report on how they moderate and respond to antisemitic content and extremism. But until we see significant change and a culture shift with regards to regulation and safeguarding, we can expect only more violence and more terror.
Steven Stalinsky is Executive Director of the Middle East Media Research Institute (MEMRI)
