With language models I won’t be surprised if it’s possible to create semantically equivalent but syntactically different text such that you can uniquely identify leakers.
Then you could send these out in a coordinated fashion to identify which group, if not person, is leaking.
The message says that they will find the leaker, but I doubt it (of course this is from 2010, so if they were found they have been for a while now).
Then you could send these out in a coordinated fashion to identify which group, if not person, is leaking.
The message says that they will find the leaker, but I doubt it (of course this is from 2010, so if they were found they have been for a while now).