CSIRO Team Confirms First Black Hole Image With New ML Technique

An Australian research team has independently confirmed the first-ever image of a black hole using a new machine-learning technique that produced a nearly identical ring structure from the original 2017 telescope data.

Researchers at Australia’s national science agency, the Commonwealth Scientific and Industrial Research Organisation (CSIRO), reanalyzed the Event Horizon Telescope (EHT) observations of the supermassive black hole M87* and published their findings in Nature on November 11, 2026. Their method, called GenDIReCT, reconstructed a ring with an angular diameter of 42.2 ± 1.5 microarcseconds, corresponding to a black hole mass of 6.5 ± 0.5 billion solar masses — matching the original 2019 EHT result.

A new algorithm, no preconceptions

The CSIRO team led by Samuel Lai developed GenDIReCT to sidestep the calibration errors and human biases that have dogged very long baseline interferometry. Unlike the EHT’s original imaging pipelines, which were tested on simulated black holes, GenDIReCT was trained on everyday images — cats, dogs, cars — so it had “no idea what a black hole looked like” when applied to the M87 data, Lai said.

When we applied it to the original EHT dataset, we found a black hole almost identical in size and shape as the historic EHT image. Importantly, our model had no idea what a black hole looked like when we applied it to the M87 data.

Samuel Lai, CSIRO astronomer

The technique uses “closure invariants” — observables immune to instrument calibration errors — and a machine-learning imaging step that was frozen after validation to prevent human bias from creeping in. The team applied it to the EHT’s 2017 observing campaign data on December 4, 2025, and recovered the ring structure across multiple data sets.

Answering the 2022 challenge

The independent check matters because the original EHT image had faced scrutiny. In 2022, a group led by Junichiro Makino at Kobe University re-ran one of the EHT's own analysis methods with slightly different assumptions and produced a blob-like structure instead of a ring, which casts doubt on whether the ring was an artifact of the imaging process.

An image of the supermassive black hole M87* revealed to the public in 2019
Photo: Space

The EHT collaboration’s original 2019 result came from four independent imaging teams that all produced a lopsided ring, but the CSIRO work is the first fully independent validation using a fundamentally different mathematical approach.

This is the first truly independent validation of the EHT image, an important step in the scientific process.

Samuel Lai, CSIRO astronomer

EHT collaboration welcomes the result

Laurent Loinard, Director of the EHT Collaboration, said the independent method recovered M87*’s ring structure across several years of data, in excellent agreement with the collaboration’s published results.

It is extremely gratifying to see that GenDIReCT, an elegant, entirely independent method, recovers the ring structure of M87* across several years, in excellent agreement with the EHT Collaboration’s published results. Studies like this strengthen our confidence in what we have learned about black holes, and enrich the toolkit available to the whole community.

Laurent Loinard, Director of the EHT Collaboration

Einstein’s prediction holds

The ring corresponds to the “shadow” of the black hole — a dark central region where light cannot escape, surrounded by superheated plasma glowing at 1.3-millimeter wavelengths. The size and shape match predictions from Albert Einstein’s 1915 general theory of relativity.

M87* sits 55 million light-years away in the galaxy Messier 87; its mass is about 6.5 billion times that of the Sun, making it larger than the orbit of Neptune. As one EHT scientist noted in 2019, an object orbiting this black hole would circle it within a week at near light speed.

Australian team confirms first black hol
Photo: UA.NEWS

Beyond black holes

The CSIRO team says GenDIReCT could transfer to other sparse-data imaging problems, including medical imaging and earthquake mapping. The technique was designed for situations where multiple sensors combine to map an unknown structure — exactly the challenge the EHT faced when stitching together eight radio telescopes across four continents into an Earth-sized virtual dish.

The study appears in Nature, published November 11, 2026.

An international team of over 200 astronomers, including scientists from MIT’s Haystack Observatory, captured the first direct images of a black hole in 2019. In a series of papers published in a special issue of Astrophysical Journal Letters, the team revealed four images of the supermassive black hole at the heart of Messier 87, a galaxy within the Virgo galaxy cluster, 55 million light years from Earth.

“This is the dawn of a new era of black hole astrophysics,” the Haystack team said.

The EHT’s angular resolution was equivalent to reading a text on a phone in New York from a sidewalk café in Paris. Even the largest radio telescopes on Earth are nowhere near big enough to see a black hole on their own, requiring the interferometric combination of multiple dishes across the globe.

Founded in 1926 as the Council for Scientific and Industrial Research, CSIRO is Australia’s national science agency, responsible for scientific research and its commercial applications. It operates across more than 50 sites in Australia, France, and the United States, employing 6,618 staff as of 2024. Among its notable achievements are the invention of early Wi-Fi technology, the first polymer banknote, and the development of biological controls to manage rabbit populations.

The article is republished from The Conversation under a Creative Commons license.

Earlier coverage: CSIRO Researchers Validate M87* Black Hole Image, Study Finds

Photo of author

Technology Desk

Technology Desk at Archyde.

U.S. Loses Measles Elimination Status Amid Over 5,400 Cases, 10 Deaths