I work at a large law firm that represents a lot of software companies. Our standard employee agreement forms have the usual default (company owns everything you create with its resources or that relates to the scope of your employment).

This default has always amused me because lawyers never sign these kinds of agreements with their own law firms. We spend most of our time writing contracts, memos, and other bits of work product that, in theory[1], are protected by copyright. Ethics rules and professional norms also give the clients rights in the work product they pay us to produce. A firm's partnership agreement might address this too. But most firms don't even try to address who owns the underlying IP rights.

Moreover, it is extremely common for partners moving between firms to take all of their forms with them. The result is that people treat contract forms as-a sort of IP-free zone. It would not even be possible to ascertain the original authorship of most form contracts that cross my desk.

Historically, this hasn't mattered because law firms charged for hours worked. It will matter a great deal if firms shift toward offering more automated products that can be sold outside the billable hour.

[1] As with source code, there's also uncertainty about which aspects of a contract are expressive and which are purely functional. Only the expressive parts are protected by copyright.

Update: one project I've had in the back of my mind is to illustrate this point by crawling the SEC's EDGAR website and tracing the "genealogy" of bits of contract language in public companies' filings. (Companies need to file certain "material agreements"). If anyone has some suggestions for good text processing libraries that can help with this tracing, I would love to hear about them.