On May 21, 2025, Elias Rodriguez approached two Israeli embassy staffers outside the Capital Jewish Museum in Washington, D.C., and murdered them in cold blood. Just hours earlier, he had posted a 900-word manifesto online that justified violent political “escalation” in the name of Gaza, framed the impending attack as legitimate protest, and called for more “armed demonstrations.”

Less than two weeks later, Mohamed Sabry Soliman, an Egyptian national living illegally in the US, set several pro-Israel demonstrators on fire near a mall in Boulder, CO. Twelve people were injured, including several elderly participants and a Holocaust survivor. Soliman later admitted to targeting “Zionist people” and claimed that he had planned the attack for over a year.

While these incidents differed in method (one attacker used a gun, the other a flamethrower) and in communication strategy (one published a manifesto, the other did not), both represent not just failures of intelligence, but failures of imagination. These attacks expose outdated methods for tracking threats, systems that fail to account for the role of radicalized language in the digital age, and social media platforms’ reluctance to share critical data necessary to detect and address these risks.

For decades, counterterrorism has focused almost exclusively on tracking networks: chatter between suspects, coordinated plots, and ties to extremist groups. These threats are real, but today’s most urgent danger comes from individuals radicalized in isolation, often online. They don’t need a group or a leader. They don’t signal affiliation or send encrypted messages. Instead, they broadcast their ideology openly.

Rodriguez and Soliman fit this pattern. Neither was a member of a known extremist group, nor did they use ciphers or communicate surreptitiously on back-channel applications. Rodriguez’s manifesto was a clear, public statement of intent to carry out an attack, while Soliman’s attack was driven by over a year of ideological hatred. Both follow a pattern seen in other lone-actor attacks like Pittsburgh (2018), Christchurch (2019), and Halle (2019). In each of these tragedies, the manifesto or incitement was discovered only after the killings. Each time, we promised to learn — but each time, we missed the same signals.

These attacks were also a byproduct of obsolete detection methods. While enormous resources go into tracking invisible networks, far too little attention is paid to what is being said online and in person — the words and sentences that reflect dangerous ideologies. Violent extremists often use dehumanizing language to justify murder, but their rhetoric is often dismissed as mere speech (and protected speech, at that).

Rodriguez’s manifesto was ignored until after the violence occurred. Soliman’s attack, while not preceded by a written screed, was the result of sustained ideological incitement. Still, each case demonstrates why we cannot continue to treat incitement as just noise. Dangerous fantasies of “resistance” and glorified violence circulate online every day, often unchecked and without consequence.

Further complicating the matter are online platforms that continue to restrict access to the very data that researchers, civil society, and policymakers rely on to monitor and prevent these threats. Under the guise of protecting privacy or free speech, they enable opacity. People are getting hurt — and dying — as a result.

Make no mistake: free speech and privacy are essential to any democracy. However, these ideals become untenable when they shield violent content, allowing it to spread unchecked. The belief that all speech is equal and non-predictive is naive. History has shown us that hate speech often precedes violence. From the Holocaust to Rwanda, we know that such rhetoric prepares the ground for action.

In a world where lone-actor violence is escalating, and incitement to violence is more openly visible than ever, the solution is multifaceted. We must change the way we listen by investing in systems and disciplines that can analyze not just slurs or buzzwords, but ideological narratives, dehumanizing metaphors, and escalating rhetoric. This includes AI capable of detecting patterns, and fields like psycholinguistics and discourse analysis, which examine radicalization as a communicative process, not just a network-based phenomenon.

We can achieve this while striking the right balance between free speech and public safety. By prioritizing the content of speech over the identity of the speaker, we can monitor threats while upholding the same legal standard we apply offline, where speech is protected until it becomes a credible, imminent threat. At that point, the state not only has the right — but the obligation — to act.

Rodriguez told us what he was going to do. His words were public, unencrypted, and visible for all to see. Soliman’s attack, though not preceded by a manifesto, was the result of ideological incitement. In each case, we failed to act.

If we continue relying on antiquated tracking methods, treating hate speech as background noise, and looking the other way as platforms restrict the very data we need to track incitement, we will remain blind to the threats emerging right in front of us. We need to listen differently, adapt our methods, and invest in the right tools before the next manifesto becomes another obituary.

Matthias J. Becker is a visiting fellow at the Tel Aviv Institute and leads Decoding Antisemitism at the University of Cambridge.

The post More Attacks Like the Embassy Murders and CO Firebombing Are Coming Unless We Change Our Anti-Terror Strategy first appeared on Algemeiner.com.