We've built an automated safeguarding layer into our Lesson transcript system. Here's how it works:
Automatic content scanning
Every Lesson transcript is automatically reviewed by an AI agent after the Lesson transcript is generated. It scans for six categories of concern, and each category comes with a plain-English description of what it covers, shown alongside the flag and in your safeguarding settings:
Profanity — swearing or crude language.
Self-harm — a participant disclosing or alluding to self-harm, suicide, abuse, neglect or exploitation.
Off-platform contact — any attempt to move communication off-platform, such as asking for personal phone numbers, social media handles, addresses or to meet in person.
Personal information — sharing of personally identifying information that shouldn't be in a Lesson, such as a home address, full date of birth or financial details.
Inappropriate content — sexual, violent, hateful or otherwise age-inappropriate content.
Bullying — a participant insulting, mocking, demeaning, threatening or pressuring another participant.
Each flag is assigned one of two severity levels, low or severe, so reviewers can prioritize accordingly.
Choosing what Bobbin looks for
Safeguarding detection is on by default, with all six categories active. An Administrator can change this under Workspace > Settings > Safeguarding:
Enable safeguarding detection — turn detection off altogether. While it's off, no flags are recorded and no alerts are sent.
Categories to detect — choose which of the six categories to look for. A concern in a category you've unticked isn't recorded and doesn't trigger an alert. Each category's description is shown next to it, so you can see what you'd be switching off.
Note: These settings affect Lessons starting from the point you change them. Flags already recorded stay on the Safeguarding dashboard.
Turning detection off, or unticking a category, is recorded in your organization's activity log along with who made the change.
Audit trail
Every concern identified is permanently recorded, creating a full audit trail. This means nothing is lost or overlooked. Even if a flag is later resolved, the record of it remains. Each flag captures the category, severity, the speaker involved and whether they were a Tutor or a Student, a relevant excerpt from the transcript, an explanation of why it was flagged, and where in the session it happened.
The speaker's role is stored on the flag at the moment it's raised, so it stays accurate even if that person is later removed from your organization.
Instant Administrator alerts
As soon as a concern is detected, an alert email is automatically sent to the appropriate Administrators. The email includes a summary of all flags from that Lesson and a direct link to review it. These alerts are sent either to owner-level Administrators or specific Administrators set to receive safeguarding alerts.
Each flag in the email names its category in full, gives the reason for the flag on its own summary line, and shows the conversation from either side of the flagged moment — roughly twenty seconds each way — with the flagged line highlighted. That's usually enough to judge whether something needs acting on before you open Bobbin.
Note: Safeguarding alerts always come from Bobbin's own email address rather than your agency's, so they reach you reliably and can't be mistaken for a message to a Client.
Letting participants know
Before anyone enters a Lesson, the join screen shows a notice explaining that the session is recorded for safeguarding purposes. Everyone joining sees it, so there's no need to explain the recording separately each time.
Review workflow
Administrators review flags through a dedicated Safeguarding dashboard. The dashboard lists flagged Lessons with their category, severity, speaker and status.
Opening a flag gives you the tools to judge it in context:
Expand the conversation. Start from the excerpt and widen the surrounding conversation to see more of what was said either side of it. Bobbin tells you when there's more of the conversation still to show, up to the maximum window.
See who was speaking. Each line shows whether the speaker was a Tutor or a Student.
Jump to the moment. Go straight to that point in the Lesson recording rather than scrubbing through it, so you can hear the tone as well as read the words.
Closing a flag
When you've finished reviewing, close the flag and record what you concluded. There are two outcomes:
Resolved — the concern was real and has been dealt with.
Not an issue — on review, the flag was a false positive.
You can add a note of up to 2,000 characters recording what you did or why you reached that view. Closed flags record the outcome, the note, and who closed them and when, maintaining full accountability throughout the process. Marking a flag as not an issue also feeds back into how the detection performs over time.
A flag can be reopened if something changes. Reopening clears the outcome and note along with the closed stamp, so the flag returns to the open list for review.
Who can see safeguarding flags
Safeguarding flags are visible to Administrators only. Tutors, Students and Clients can't see the Safeguarding dashboard or any flag raised against a Lesson, including their own.
Where a flag links through to a Lesson recording or transcript, access to that recording and transcript follows your organization's recording visibility setting, so safeguarding review doesn't open up recordings more widely than you've chosen.
Recordings linked to open flags
If your organization has set a retention period for Lesson recordings, a recording connected to an unresolved safeguarding flag is kept rather than removed on schedule, so the evidence is still there when you come to review it. Once the flag is closed, the recording follows your normal retention period again.
Note: On Bobbin Lite?
Safeguarding detection is part of Bobbin's AI features, so it isn't included on Bobbin Lite and no new flags are raised.