> The key point here is that the shape of the glyphs themselves, for example, non-trademarked text posted on advertisements or products with printed text, are not copyrightable. So long as you’re not stealing the creative work, like the advertisement, itself, the shape of each letter is in the public domain.
There is a jump from this assertion to:
> This got me thinking, as all things do, on whether I could simply scrape the internet for public, non-creative, non-trademarked use of fonts
The missing step here is that the glyphs for a webpage are not embedded in some sort of rendered image or physical medium. They're _rendered from the copyrighted font file_ to the browser.
The font file is delivered to the browser and that copyrighted file is used to render the glyphs.
This pipeline might work for creating a font that is "inspired by" the available glyphs rendered in the credits of your favorite film.
But downloading the entire copyrighted font file, generating a character sheet from it and clipping the individual characters from that sheet? I can't imagine how any sensible court system would not consider that a derivative work.