O'Neill reiterates that other forms of access to notes will only occur in "very limited cases" where there's no other choice. Evernote may have to comply with a warrant, for example, or investigate claims of harmful content or technical issues. There's only an "extremely limited" number of people allowed to look, and O'Neill helps vet them.
The defense is slightly clearer and more reassuring than before, but it ultimately amounts to a repetition of what the company already said: trust us, we'll do our best to avoid reading anything you don't want us to see. It doesn't really address some of the outstanding questions, however. How do you know that the machine learning system will always scrub identifying data? What about confidential business info that's unlikely to be cut? And when Evernote staffers have to read info for reports of abuse or troubleshooting, how do you know that they won't read more than they have to? It's doubtful that Evernote will abuse its power when the policy takes effect on January 23rd, but that won't be completely comforting if you want a guarantee that the company won't mishandle your sensitive material.