The WIV had the largest program in the world to sample novel sarbecoviruses from nature. At the beginning of the pandemic, the published virus closest to SARS-CoV-2 (RaTG13) was from the WIV. Closer viruses (BANAL) have since been published, by a different group but from areas where the WIV was also recently sampling.

There's no serious question that the WIV has unpublished viruses--even with no attempt at secrecy, every active research group has unpublished work. Researchers found an unpublished merbecovirus in contamination from shared equipment. This isn't related to SARS-CoV-2, but shows the claim that the WIV had zero unpublished viruses to be specifically false. Public access to the WIV's database of viral genomes was removed early in the pandemic, and never restored.

https://www.biorxiv.org/content/10.1101/2023.02.12.528210v2

Pangolins were initially proposed as the proximal host, but that's been abandoned for years. After a long delay, the paper in Nature was extensively corrected, following Alina Chan's discovery that the alleged multiple samples were all from a single batch of smuggled pangolins. These were probably infected during trafficking, in the same way that housecats are sometimes infected by SARS-CoV-2 but aren't the source.

https://www.nature.com/articles/s41586-020-2313-x

https://www.biorxiv.org/content/10.1101/2020.07.07.184374v2

The goal of research like DEFUSE was gain of function, a deadlier or faster-spreading virus. That goal wasn't always successfully achieved, but that's true for all goals. The point is that skilled researchers specifically trying to achieve a goal (like by directed evolution during serial passage, or by genetic engineering) are much more likely to do so than would random point mutations alone.

None of this means it's certain that SARS-CoV-2 arose from an accident at the WIV. The picture that you've received isn't accurate, though.