AI tool reveals hundreds of ‘anomalies’ in Hubble telescope archive – some defy classification

Artificial intelligence (AI) tools have discovered more than 1,000 strange cosmic objects in the Hubble Space Telescope image archive, some of which cannot be explained by science.

After just two days of searching with the tool, researchers discovered 1,300 strange objects, including chaotic merging galaxies, gas-trawl stars, and even objects that have yet to be classified. Of these, 800 have never been discovered before, European Space Agency (ESA) officials said in a statement on January 27. The results of this research were published in the academic journal Astronomy & Astrophysics on December 16, 2025.

“Additional discoveries include galaxies with massive star-forming clumps, jellyfish-like galaxies with gaseous ‘tentacles,’ and a planet-forming disk at the edge of our galaxy that resembles a hamburger,” NASA officials said in a separate statement. “Surprisingly, dozens of objects completely defied existing classification schemes.”

you may like

space jelly and sky burger

For the new study, ESA researchers David O’Ryan and Pablo Gomez developed an AI tool that examined 100 million image clippings from the Hubble Legacy Archive, which covers the telescope’s observations since its launch in 1990. Each image is just a few dozen pixels on a side and represents a narrow slice of the sky, just 1/1000th of a degree wide.

“Archive observations from the Hubble Space Telescope, which now date back 35 years, represent a treasure trove of data from which astrophysical anomalies can be discovered,” O’Ryan said in the paper.

In addition to the “jellyfish galaxy” and the cosmic “hamburger”, the survey also discovered various other phenomena. “Most of the anomalies are galaxies merging or interacting, exhibiting unusual morphologies and elongated streams of stars and gas,” NASA said in a statement. “Others are gravitational lenses, where the gravity of foreground galaxies distorts space-time, bending light from background galaxies into arcs or rings.”

The researchers’ AI tool, called AnomalyMatch, detected these features after learning patterns from a training dataset. Such tools speed up traditional methods of spotting strange things in the sky, which usually require manual inspection or lucky observations.

“While professional astronomers are adept at identifying unusual features, the sheer volume of Hubble data makes a comprehensive manual examination impractical,” NASA officials said in a statement. “Citizen science efforts have helped expand the scope of data analysis, but these efforts fall short in the face of extensive archives like Hubble.”

“This is a powerful demonstration of how AI can enhance the scientific return of archival datasets,” Gomez added. “The discovery of many previously undocumented anomalies in the Hubble data highlights the potential of this tool for future investigations.”

Other datasets where AI could be useful include those from the Euclid Space Telescope, which is surveying billions of galaxies to create the largest 3D map of the universe in history, and the Nancy Grace Roman Telescope and Vera C. Rubin Observatory, which will search for exoplanets and moving objects across vast swaths of the night sky. AI could help researchers sort through “large amounts of data” from these large-scale surveys, possibly allowing them to discover new objects faster than ever before, according to a statement from NASA.