European Space Agency scientists have deployed an artificial intelligence system to comb through decades of NASA’s Hubble Space Telescope data, uncovering more than 1,300 unusual cosmic objects — including over 800 that have never been documented before — by rapidly scanning nearly 100 million archived image cutouts, a task that would take human researchers far longer to complete and showcasing how AI can boost scientific discoveries even amid budget constraints for space agencies. According to the reporting, the ESA-developed neural network, called AnomalyMatch, flagged patterns and rare phenomena in the massive Hubble archive in under three days, with findings ranging from merging galaxies and gravitational lenses to jellyfish-shaped galaxies and objects that defy current classification, indicating that even long-studied datasets still hold hidden scientific value when paired with modern machine learning tools.
Sources:
https://www.semafor.com/article/01/28/2026/european-space-agency-uses-ai-to-identify-nasas-hubble-telescope-anomalies
https://www.esa.int/Science_Exploration/Space_Science/1400_quirky_objects_found_in_Hubble_s_archive
https://phys.org/news/2026-01-ai-hundreds-cosmic-anomalies-hubble.html
Key Takeaways
- European Space Agency researchers developed a neural-network AI called AnomalyMatch to analyze nearly 100 million Hubble image excerpts and find rare astronomical phenomena.
- The AI discovered over 1,300 anomalous cosmic objects, more than 800 of which had never been documented in scientific literature, revealing previously missed insights in long-standing datasets.
- The rapid AI analysis dramatically outpaced what manual review by human scientists could achieve, underscoring how machine learning tools can expand scientific discovery even when traditional funding pressures challenge space research efforts.
In-Depth
In a development that highlights both the power of artificial intelligence and the enduring scientific value of existing space data, researchers from the European Space Agency have successfully applied AI techniques to uncover a trove of unexpected cosmic phenomena hidden within the archival images of NASA’s Hubble Space Telescope. The effort centers on a custom neural network dubbed AnomalyMatch, designed to sift through vast amounts of image data — nearly 100 million Hubble image cutouts — and identify those that exhibit unusual or previously unseen features. Rather than relying on the painstaking manual review that has traditionally accompanied such work, this AI-driven approach processed the immense dataset in just about two and a half days, flagging more than 1,300 potential anomalies, with over 800 of those lacking prior recognition in scientific databases.
The range of discoveries is remarkable in its diversity. Many of the flagged objects are galaxies in the throes of dramatic interactions: merging systems with distorted shapes and extended tidal tails, rare gravitational lens effects where a foreground galaxy’s mass bends and warps the light from objects behind it, and ring galaxies shaped by previous collisions. The findings also include striking examples like jellyfish galaxies, whose gaseous streams give them a tentacled appearance, and planetary-disk structures seen edge-on that resemble unexpected shapes. Perhaps most tantalizing are the anomalies that resist easy categorization — objects that don’t fit neatly into established astrophysical classifications and suggest that even well-studied portions of the universe can surprise scientists when examined with fresh tools.
Crucially, these results underscore how AI can amplify scientific output without the need for extensive new missions or costly hardware. With space science funding, particularly at NASA, facing fiscal pressures and tough budget choices, tools like AnomalyMatch help maximize the scientific return on existing investments. Rather than letting enormous datasets languish because they’re too large for human researchers to process in detail, AI enables a thorough, systematic review that reveals hidden gems and broadens our understanding of the cosmos. This approach also sets the stage for future work, not only with Hubble data but with the even larger datasets that next-generation telescopes will produce. By blending decades-old observations with cutting-edge machine learning, researchers are rewriting what’s possible in astronomical discovery — without having to wait for the next launch window.