Redwood City, Calif.-based Evernote Corp. amended its policy in anticipation of deploying machine-learning capabilities that, according to a company FAQ page, will allow the app service to automate certain functions, such as the creation of to-do lists and itineraries. The company is promising that the functionality will create more a convenient, personalized experience. But there is a trade-off: according to Evernote, its data scientists must perform spot checks on the technology to make sure it is processing information correctly – and this requires access to user content.
“While our computer systems do a pretty good job, sometimes a limited amount of human review is simply unavoidable in order to make sure everything is working exactly as it should,” reads the FAQ page.
Users have been given the choice to opt out of the machine-learning program. However, in making its announcement, Evernote also drew attention to several conditions in its current policy that already give employees the right to review content under additional circumstances. In an update to its FAQ page, the company clarified that such exceptional circumstances include when a warrant is served, when harmful or illegal content is reported, and when troubleshooting the service at the request of a user.
In Evernote's online forums, some users have expressed dismay over the forthcoming changes to the policy – in several cases threatening to abandon the service.
In a company blog post on Thursday, Evernote CEO Chris O'Neill sought to ease fears over how app users' content will be treated by data scientists working on the machine-learning technology. “Select Evernote employees may see random content to ensure the features are working properly, but they won't know who it belongs to,” wrote O'Neill. “They'll only see the snippet they're checking. Not only that, but if a machine identifies any personal information, it will mask it from the employee.”
The app developer noted that it will still make machine-learning technologies available to users, "but no employees will be reading note content as part of this process unless users opt in." Under the rejected policy, users would have had to opt out of the program.