SaneBox calls its decade-old header sorting machine learning

The setup

SaneBox has sorted incoming mail into folders like SaneLater since long before the current AI boom, working by connecting to a mailbox and adding server-side folders rather than replacing the email client itself. SaneBox's own FAQ page states the service establishes a secure connection to the mail server, analyzes only header data such as sender and subject rather than the body of a message, and moves lower-priority mail into its folders before a person ever opens them, a mechanism the company now markets under the label SaneBox AI on its homepage.

What the documents show

The FAQ states plainly that SaneBox uses sophisticated algorithms and machine learning to automatically sort incoming email by importance, so the filtering mechanism itself is described by the vendor as machine-learning based, not a plain rules engine with AI bolted on afterward. Layered on top of that sorting, SaneBox's Daily Digest guide describes a daily review screen that groups filtered emails by folder, shows a snapshot of time saved and emails received, and lets a person move, train, or archive a message without opening it. The guide describes this Digest as interactive rather than a generated narrative summary, distinct from a block of AI-written prose.

The friction

The FAQ discloses a specific privacy trade-off behind the sorting mechanism: SaneBox never reads the body of an email, only header fields, which limits how much context the underlying model has to work with when deciding importance, and the company states it never stores full emails or attachments as part of that same design choice. The Daily Digest guide does not describe an AI-generated one-line summary of what an individual email says, so the marketing label of an AI-powered inbox assistant covers header-based classification more clearly than it covers narrative summarization of message content.

What changed in the work

The documented change predates generative AI: a person stopped manually filing newsletters and low-priority senders years before large language models existed, and the Daily Digest simply gathers that filtered mail into one interactive review screen instead of several folders checked separately. Calling the current version SaneBox AI is accurate to the vendor's own machine-learning claim for sorting, but reading a generative, content-aware summary into the Digest itself would go beyond what the guide describes.

• Does the header-only analysis miss context that would change how an email gets sorted, given SaneBox states it never reads message bodies?
• Is the value being attributed to AI actually coming from the years-old sorting mechanism or from the newer Daily Digest review screen?
• What does moving a message back to the inbox actually teach the sorting model, according to the training instructions in the FAQ?

SaneBox's own FAQ credits machine learning for the decade-old act of sorting mail, while the newer Daily Digest is documented as an interactive list rather than an AI-written narrative, a distinction worth keeping straight before assuming every AI-branded feature here is new. The friction is less about consent or cost and more about which part of the product the AI label actually describes.