TheRegister's platform is a leading technology news website, offering insights into IT industry news, hardware reviews, and software updates. Through articles, analysis, and opinion pieces, TheRegister offers insights into cybersecurity threats, technology trends, and industry developments. Readers can stay updated with the latest news and analysis from the world of technology and IT business.

The Register

Global technology corporations often ignore laws in their pursuit of wealth, facing minimal consequences like fines. The post argues for making illegally trained large language models (LLMs) public domain to deter unethical practices, considering the high environmental cost of simply deleting them. It emphasizes the need for stricter regulations and global cooperation to ensure companies comply with legal obligations related to privacy and data protection.

Make illegally trained LLMs public domain as punishment

So LLM based on a illegaly obtained private photos of one specific person, trained to build porn with every query should be just become public?
Well I see some problems with this appraoch.

That’s kind of like saying “make stolen cars public property”. Normally, you’d return the property to the original owner. However, in this case the “property” cannot be returned, nor can it be purged from the source. Licensing also compounds this issue. If your data was there and you’re OK with it being offered “open source” but not commercially, that doesn’t mean everyone else is. If the LLM was trained on “illegal” data, then the only recourse is the destruction of the LLM.