MNPI detection in call transcripts

The short answer

Junior screens every call for potential material non-public information against rules compliance sets once, so compliance reviews findings, not calls.

A finding is a prompt for human review. It shows the passage, the rule and the reasoning, and the decision on whether a passage is material and non-public stays with the firm’s compliance team and counsel.

MNPI in a call context

Material non-public information is, in general terms, information about a company that has not been made public and that a reasonable investor would be likely to consider important.

Deal teams meet it routinely, in management meetings, expert interviews and banker calls, and a transcript turns a spoken remark into a durable record. Gunderson Dettmer’s June 2025 presentation for fund managers lists recordkeeping and discovery risk among the main risks of AI note takers for that reason.

This page is general information and does not constitute legal advice.

Layers of control

A firm can control MNPI exposure at three points in the life of a call, and a note taker contributes to each of them.

How Junior detects MNPI

Compliance sets the rules once, and Junior screens every passage of every call against them. MNPI is checked against public sources, so news that is already out isn’t escalated.

Compliance reviews findings, not calls. Each finding shows the passage, the rule and the reasoning, and can be dismissed, confirmed or escalated, with a default action per rule. Open findings stay hidden from the deal team until compliance acts.

Ethical walls keep calls inside the project unless policy allows otherwise, and every flag, decision and change is recorded with a timestamp in the audit trail. Junior was built with advice from SEC-grade counsel, and it lets firms that had banned transcription record calls for the first time.

The basics sit underneath. Junior is SOC 2 Type II certified, never uses calls to train a model, has zero data retention agreements with every underlying model provider, and keeps call data on its own infrastructure.

What other vendors publish

Other AI note takers describe their own MNPI features, and a compliance team should compare them on the same questions rather than on the label.

Fellow’s pages describe AI trackers for potential MNPI or PII in a compliance portal add-on, the ability to pause and resume recording during sensitive discussions, and transcript redaction. Those are Fellow’s own descriptions, reviewed on 11 November 2026, and they are worth testing on sample calls like any other vendor claim.

Questions to ask before relying on detection

Five questions turn a feature claim into something a compliance team can evaluate.

• What does the model flag, and how is it trained for the calls our firm actually runs?
• Who sees a flag, and what record is kept of the review?
• What happens to a flagged passage in the transcript, the summary and any exports?
• Can retention differ for audio, transcript and summary?
• Can the firm test the feature on its own sample calls before signing?

Common questions