On February 26, 2026 – 48 hours before the United States and Israel launched Operation Epic Fury – Hangzhou-based MizarVision posted AI-annotated satellite imagery of F-22 Raptors at Ovda Air Base in Israel, air strike group positions in the Arabian Sea and AWACS aircraft at Prince Sultan Air Base in Saudi Arabia.
The company does not own satellites. It acquires commercially available imagery from Western and non-Western providers and processes it through proprietary AI that identifies and categorizes military assets through automated object detection.
Operation Epic Fury began approximately 24 hours after a release of particularly significant images. MizarVision continued posting throughout the campaign, including damage assessments following the Iranian strikes on Al Udeid, the repositioning of the THAAD battery in Jordan, and the locations of tanker aircraft at bases across Bahrain and the United Arab Emirates. The US military classified many of its force movements. A Chinese AI trading platform published them on social media in near real time.
Conventional understanding of operational security in the satellite era assumed that the relevant limitation was satellite access. Cold War spy satellites were large, few, expensive, and operated by states, meaning that only major powers had the continuous surveillance capability to track the movements of military forces with significant time resolution.
Commercial satellite constellations changed this calculation over the past decade, but the full implications were not operationally demonstrated until Epic Fury. Planet Labs ACTS more than 200 SmallSat imaging satellites with daily review capability over any point on Earth.
The Maxar constellation provides sub-meter resolution imagery available by commercial subscription. These are not classified systems – they are commercial products accessible to any buyer with the right subscription and computing capacity to process images at scale.
What MizarVision demonstrated is that commercial image processing with AI-enabled object detection compresses the timeline from satellite transition to military intelligence product from days to hours.
Force behavior information that once required a classified satellite program and a dedicated processing facility on the ground can now be extracted from commercial imagery by a private company that posts its findings on social media.
The operational consequences of MizarVision’s activities during the campaign were significant. The F-22 Raptors at Ovda were identified by a satellite pass and their presence was publicly confirmed before the start of the operation.
THAAD battery repositioning in Jordan and Saudi Arabia was tracked through back-to-back passes, revealing defensive configuration changes that influenced Iranian retaliation planning. The AWACS positions on the Prince Sultan were released, and the Prince Sultan was subsequently hit repeatedly by Iranian attacks.
Technical OSINT geolocated Destroyed E-3 Sentry after its loss, identifying it as serial number 81-0005 from available aircraft markings and satellite imagery. Whether MizarVision’s releases directly informed Iranian targeting—and whether the correlation between the release of the images and subsequent strikes reflects causal influence—remains unresolved from public evidence.
The operational reality that the correlation documents is analytically significant, regardless of causation: US force posture in seven countries was consistently visible to any actor who had access to MizarVision’s social media posts throughout the campaign.
The governance architecture for commercial satellite imagery and its military applications does not address this reality. The US government operates a system of shutter controls – the authority to restrict commercial image providers from releasing images of specific locations during sensitive operations – over US companies.
Planet Labs and Maxar are subject to this authority. MizarVision is not.
It’s a Chinese company that buys images from non-US providers, processes them with AI and shares them on social media platforms. There is no international agreement governing the military use of commercial satellite imagery by non-state actors, and there is no norm restricting the release of AI-processed military force posture data derived from commercial imagery.
The Arms Export Control Act and International Arms Traffic Regulations regulate the export of military systems and technology, not commercial images that allow adversaries to track those systems.
The strategic implication is not primarily about MizarVision specifically, or even China’s use of commercial intelligence more broadly. It is about the structural transformation of the surveillance commons.
Every military in the world is now conducting operations under the assumption—confirmed at an operational scale for the first time—that hiding force posture from commercial satellite surveillance is not achievable for assets visible from orbit.
This includes not only large platforms such as aircraft carriers and AWACS aircraft, but increasingly smaller and harder-to-hit targets as AI object detection improves.
The operational response—US Air Force images of the destroyed E-3 were not released, commercial images from Prince Sultan’s American providers were delayed, and assets were dispersed across isolated connecting routes to complicate targeting—reflects the adaptation of force protection doctrine to a surveillance environment that did not exist when the doctrine was written.
The governance gap is real and cannot currently be addressed through unilateral US policy. An international agreement limiting the military use of commercial satellite imagery by non-state actors would require the participation of China — the country whose commercial intelligence companies most directly exploited the capability during Epic Fury — and would represent the most important arms control negotiations since the Open Skies Treaty, from which the US itself withdrew in 2020.
The commercial satellite commons will be ungoverned indefinitely. The question is whether the government exits the negotiations before the next conflict demonstrates more serious consequences than the visibility of the force posture revealed in 2026.
Aleena Saif Ullah is a scholar in international relations specializing in global defense and security at the University of the Punjab, Lahore.
