Did U.S. Airstrikes in Yemen Use Open-Source Intelligence To Target Civilians?

On April 27, 2025, a powerful explosion rocked the outskirts of Yemen’s Sanaa province. The blast was the result of a U.S. airstrike targeting an isolated site initially believed to be a military facility, part of an ongoing campaign against Houthi positions. However, it was later revealed that the strike had hit a civilian location.

As images from the scene spread online, questions quickly surfaced about the nature of the target. Satellite imagery later showed that the site had no identifiable military features. What drew particular attention was that the strike came just days after a well-known open-source intelligence (OSINT) account on X, named Vleckie with the handle @VleckieHond, identified the same location as a Houthi military site. The claim was based on a visual analysis of the area’s terrain and nearby structures.

The incident has renewed debate over the role of open-source intelligence in modern warfare and the risks posed by unverified, sensitive information shared by non-professionals — data that could end up guiding military actions toward civilian areas.

How Open-Source Misinformation Turns Faulty Analysis Into Airstrikes

In recent years, especially following Russia’s invasion of Ukraine, open-source intelligence (OSINT) accounts on social media have increasingly functioned like alternative military newsrooms. These accounts provide real-time analysis based on satellite imagery, visual content breakdowns, and conflict zone maps. One such prominent account is Vleckie, which focuses on Yemeni affairs and monitoring Houthi movements.

In this instance, Vleckie shared what it claimed was a detailed analysis of a potential Houthi military site in the Thuqban area of the Bani Al-Harith district, on the outskirts of Sanaa. The account said its assessment relied in part on a post by Galal Alsalahi on X, published on April 1, which claimed to reveal a Houthi missile launch site over Sanaa.

The analysis quickly gained traction, with other OSINT-focused accounts amplifying the claim. Just days later, on April 27, U.S. forces carried out an airstrike on the same coordinates identified in the post.

The strike killed 12 civilians, including six children from one family — that of Ali Yahya Salah Masoud — according to residents and people close to the family.

In addition to the civilian casualties, inconsistencies emerged. Satellite images and local accounts confirmed that the targeted site was entirely civilian, with no visible signs of military activity. March satellite images showed a residential extension of the Thuqban Bani Al-Harith village, where trucks and excavation equipment had been clearing illegal building structures as part of a demolition campaign launched by district authorities in January.

The location also lies near warehouses operated by the World Food Programme. Low-quality satellite images showed traces of the strike. To verify the damage, analysts filtered out vegetation using a false color index, since plants often appear brown in low-resolution imagery, making it difficult to distinguish them from soil disturbances. After isolating the vegetation and comparing images of the site, clear changes were visible after the strike.

Since March, the United States has escalated its air campaign in Yemen under Operation Rough Rider. The strikes, which began on March 15, targeted what U.S. officials described as military infrastructure used by the Houthis to attack commercial and military vessels in the Red Sea and Gulf of Aden, in the context of the Gaza war that erupted in October 2023.

By April, U.S. Central Command (USCENTCOM) reported striking more than 1,000 targets, killing hundreds and reducing the Houthis’ attack frequency. On May 6, the United States and the Houthis agreed to a cease-fire, brokered by Omani mediation.

However, the U.S. air campaign also caused civilian casualties. The Thuqban incident was not isolated. The following day, a U.S. strike hit a detention center for African migrants in Saada, killing about 70 migrants and injuring dozens more. The United States acknowledged it was aware of the incident but stopped short of directly confirming involvement. However, munition fragments recovered from the site identified the weapon as a U.S.-manufactured GBU-39 bomb.

This was not the first time U.S. forces were suspected of relying on social media posts to identify potential targets. During operations against ISIS in Iraq and Syria, retired U.S. Air Force Gen. Herbert Carlisle revealed that intelligence and surveillance teams monitored social media for potential leads. In one case in 2015, U.S. warplanes reportedly carried out a strike on a suspected ISIS location based on social media content.

Though those strikes were intended to hit militant positions, they also resulted in civilian casualties, which U.S. Central Command described as “unintended losses.”

While it might seem reassuring to believe that U.S. military operations rely exclusively on high-level intelligence and precision targeting, cases like these tell a different story. They reveal a partial reliance on social media posts — sometimes from unverified accounts — in gathering intelligence and even determining potential military targets.

Although @VleckieHond later apologized for the error and insisted that U.S. Central Command does not take targeting data from Twitter, Gen. Carlisle’s earlier account of the 2015 strike contradicts that claim.

Other evidence suggests that certain levels within the U.S. military might have been aware of Vleckie’s analyses, especially those about Yemen. In April 2024, a report by the Combating Terrorism Center (CTC) at West Point cited Vleckie’s work on military operations in Yemen, describing the account as a “skilled analyst.”

The Combating Terrorism Center at West Point is part of the United States Military Academy, a military college that trains U.S. Army officers. Graduates earn a bachelor’s degree in military science and the rank of second lieutenant.

Confirmation Bias and the Analysis of Competing Hypotheses

One factor likely contributing to the mistake made by the @VleckieHond account is confirmation bias—the tendency to favor information that supports existing beliefs while ignoring or downplaying contradictory evidence. In this case, Vleckie relied on a post by social media activist Jalal Alsalahi (@galal_alsalahi), originally from Yemen’s Ibb province and currently living in the United States. Alsalahi is known for his strong anti-Houthi stance but does not appear to have verified journalistic credentials or formal training in open-source investigations.

Despite this, Vleckie accepted Alsalahi’s information, seemingly influenced by shared biases, as there was no serious consideration of alternative explanations. For example, Vleckie did not compare the targeted site in Thuqban Bani Harith with other confirmed Houthi military locations, nor did he consider that the excavation equipment visible in satellite images could be used for civilian purposes, such as removing illegal structures, as previously noted.

To avoid relying on merely satisfactory explanations—those that simply confirm the investigator’s expectations without properly testing the evidence—analysts use the competing hypotheses method. This approach requires proposing all plausible explanations, gathering all relevant evidence without exclusion, evaluating source credibility, and testing each hypothesis against the strongest evidence. The most likely hypothesis is the one with the fewest contradictions.

Applying this method to the bombing in Thuqban reveals two main hypotheses: the site is either a Houthi military base or a civilian area. Satellite images could partly support the first hypothesis because of some structures that might initially appear suspicious. Alsalahi’s post seems to reinforce this view. However, relying solely on these points reflects confirmation bias, especially when ignoring evidence supporting the alternative explanation.

Conversely, multiple pieces of evidence support the civilian-site hypothesis. Local reports describe ongoing efforts to remove illegal buildings in the area, explaining the presence of excavation and transport equipment. A time-lapse analysis of satellite images shows a concrete foundation that later disappeared, consistent with demolition of unauthorized structures. The equipment seen is primarily for civilian use, with no clear military hardware. Finally, Alsalahi’s evident anti-Houthi bias weakens the objectivity of his claims. Together, these factors make the civilian hypothesis more consistent and less contradictory with the evidence.

Read More

UAE, Saudi Arabia Deny Involvement in U.S. Talks on Yemen Land Offensive

The Invisible War: Cyber Warfare, Malinformation, and the Erosion of Privacy