> the publishers never gave permission to use their contentYes they did, yes they did. The publisher explicitly granted access by putting up the news content in an public subdirectory of a public webserver. News items don't just magically `pop up' in the public subdirectory of a webserver. Usually at least two employees (writer, reviewer) have to log into, and explicitly perform several steps in an CMS to publish a news item.
The decision of publisher is encoded as HTTP server configuration (including CMS, if any). The server obviously provides access only to some resources, as the operator wishes -- the ones meant for public consumption, while other resources are password-protected, only accessible from certain network addresses, or not accessible over HTTP at all.
Just as much as an image resource can be, and often is, protected from deep-linking based on HTTP `Referer' header, a news item can be protected from being indexed by Google News bot and displayed as a snipped in Google News based on `User-Agent' header.
In more details:
A typical User Agent that HTTP GETs some resources from the server describes itself with a `User-Agent' header. Google is honest here, and openly indicates its user agent as a spider bot, with explicit purpose of indexing content for web search.
Even better, Google uses a /separate/ spider bot for Google News service, and it is entirely reasonable to serve the general websearch bot with content -- so the website itself remains indexed -- and deny the news bot access, to prevent Google News from displaying news snippets.
Upon HTTP request, the server reads through request, including resource URL, headers (including User-Agent and Cookies, if any), compares that info to configuration. If, and only if, it matches configuration for unrestricted access, the resource is served. Otherwise, access denied condition is indicated and optionally authentication requested.
Let's assume for a moment the law goes through as requested by the press. The most probable implementation would be, Google includes a particular header in HTTP requests -- indicating itself as a `news-bot-that-has-subscription', and the server would, in turn, compare the header with configuration and either reply with news contents, or perhaps indicate access denied, should subscription expire, or for some other reason. The mechanics is pretty much same as it would be today to restrict Google News bot from gathering, and Google News from displaying (some or all) news snippets.
In short, the publishers have already configured some resources as protected from /some/ User Agents (based on request URL, headers, cookies, network address or anything else), and it is entirely their own choice they serve the news to both Google bots (websearch and news).