Content Spam Scoring

SpamCheck.ai returns a spam probability for a piece of text. When a new comment, form entry, or message arrives, a Beam agent submits the content, reads the returned score, and applies the customer-approved cutoff. Entries below the spam threshold pass through and the submitter proceeds normally. Entries above it are quarantined and the moderation queue is notified. Anything landing in a gray band near the cutoff, where the model is unsure, is set aside for a human moderator rather than auto-rejected, so genuine users are not silently blocked by a single borderline call.

Submission Quarantine

Flagged submissions need somewhere to sit before a decision. Once SpamCheck.ai marks an item as likely spam, the agent moves that record into a holding state, tags it with the score and reason, and notifies the team responsible for that channel. Repeat offenders matching an approved block rule can be held automatically. First-time senders or messages with mixed signals are handed to a moderator with the full context attached, so the team keeps a clear audit trail and no legitimate first contact is discarded without a person seeing it.

Sender Reputation Checks

Patterns across submissions hint at coordinated spam. A Beam agent reads scores over a rolling window for each sender or source, applies an approved rule on frequency and average score, and updates a reputation note on the record. Sources that cross the agreed line get flagged for throttling and the owner is alerted. Borderline sources, or ones with a sudden change that could be a false alarm, are routed to a person before any block is applied, keeping enforcement grounded in a decision someone has reviewed rather than a raw tally.