Is Plagiarism a Stain on Open-Source AI?

The open-source artificial intelligence community has been buzzing over allegations that three U.S.-based developers—including two Stanford University undergraduates—copied parts of a multimodal large-language model developed by a Chinese lab founded by Tsinghua University and Beijing-based AI startup ModelBest.

This incident might seem trivial to many onlookers, especially compared with higher-profile examples of developers ripping off AI models. But it has struck a chord with the industry—perhaps because there are fewer known instances of people copying multimodal LLMs, also known as vision language models, or perhaps because of how the developers accused of plagiarism handled the allegations.

“This might have been the week-end where blind trust in [the] VLM community kinda died?” Lucas Beyer, a researcher at Google DeepMind, said in a post on X. Beyer worked on SigLIP, one of the models the Chinese lab, OpenBMB, used to build its own.