How Intelligence works: AI, Bayes, and reputation

The Intelligence page lets you control the advanced spam detection features that work alongside Cleanbox’s standard filtering. These features learn and improve over time based on your feedback and community data.

Accessing Intelligence

Click Intelligence in the sidebar (the brain icon). You’ll see a page with up to four modules, each with a toggle to enable or disable it.

The four modules

AI Protection

AI Protection uses artificial intelligence to analyze email content and detect spam that traditional rules miss. Unlike pattern-based filters, the AI understands context — it can tell the difference between a legitimate order confirmation and a phishing attempt that looks like one.

• When enabled, every incoming message is analyzed by the AI classifier in addition to standard spam rules.
• AI Protection is available on plans that include AI features. If your plan doesn’t include it, you’ll see an option to upgrade.

Adaptive Learning (Bayes)

Adaptive Learning builds a personalized model of what spam looks like for your account. It learns from your feedback — every time you mark a message as spam or not spam, the system gets a little smarter.

• The more feedback you give, the more accurate it becomes.
• Toggle this on to enable Bayesian learning from your actions.

Tip: get in the habit of marking false positives (legitimate mail in quarantine) and false negatives (spam that slipped through). This is the single best way to improve your filtering over time.

Reputation

Reputation leverages community data from across all Cleanbox users to identify known spam senders.

• If many Cleanbox users block or report a sender, that sender’s reputation score goes down.
• Conversely, senders that are widely whitelisted get a trust boost.
• Toggle this on to benefit from crowd-sourced sender reputation.

Auto-Manage

Auto-Manage appears when AI Protection is enabled. It automatically updates contact states based on accumulated feedback. This feature is covered in detail in a separate article.

How they work together

These features add to your standard spam detection — they don’t replace it. When a message arrives, it is scored by standard rules and any Intelligence features you’ve enabled. The combined score determines whether the message is delivered, quarantined, or rejected.

Think of it as layers of defense:

• Standard rules catch known spam patterns.
• AI Protection catches novel or sophisticated spam.
• Adaptive Learning tailors filtering to your specific preferences.
• Reputation draws on community intelligence to flag bad actors early.

Recommendation

For most users, enabling all three — AI Protection, Adaptive Learning, and Reputation — gives the best protection. They complement each other: AI catches novel spam, Bayes learns your preferences, and Reputation leverages the collective experience of the Cleanbox community.

You can always toggle individual features off if you prefer a more minimal setup. Your settings take effect immediately — no restart or waiting period required.