The Famous Black Hole Picture Holds up When Another Team Rebuilds It

Writing in Nature on Nov. 11, 2026, researchers at CSIRO, Australia's national science agency, report that they rebuilt the Event Horizon Telescope's picture of the black hole at the center of the galaxy M87 with an imaging method that assumes nothing in advance about the shape it should find, and the ring came back. They put its angular diameter at 42.2 ± 1.5 microarcseconds.

The doubt is why the work was done. The ring was first published in 2019 from sparse radio data, and outside groups have since argued that it owes more to the choices made while building the image than to the sky; at least one reported that a single elongated bright spot fits the data better once the field of view is widened. Samuel Lai and colleagues say they set out to resolve that tension, and describe what they got as an independent verification of the ring geometry.

Their algorithm, GenDIReCT, is a machine-learning image generator trained on pictures that have nothing to do with astronomy. Rather than calibrated telescope readings, it is fed closure invariants, combinations of the array's measurements that are unaffected by calibration errors. The team ran it over every public Event Horizon Telescope dataset of M87 from 2017 through 2021, at several field-of-view settings, and reports a compact ring of steady size and orientation in each case.

From that diameter the authors infer a black hole mass of 6.5 ± 0.5 billion solar masses, assuming nominal scaling parameters, the standard conversion from ring size to mass. The picture best matches a system viewed about 20 degrees from the approaching jet. Their analysis does not pin down how fast the black hole spins.

The telescope data they used were already public, and the authors say their code and their reconstructed images are public too.