The best open weight models are more that capable to do this. The restrictions are probably being put in place bby Anthropic & OpenAI to have plausible deniability for possibly incoming lawsuits?

If your training is based on downloading & ingesting anything and everything w/o consent you presented evidence that respecting intellectual property is not something you genuinely care about.

So why should your paying (!) users ... unless you'd get any fallout from that.

Original title "'Reverse Engineer Anything': AI Companies Crack Down on the Coolest Thing People Were Doing With AI" compressed to fit within title limits.

(title not editorialized in an attempt to keep with guidelines, please update the title as appropriate)