Expert networks now run compliance as a three-stage pipeline on every engagement: pre-call screening, in-call moderation, and post-call transcript review. The shift was forced by the 2023 Capvision raid in Shanghai and a wave of SEC enforcement actions against platforms in 2024, and it is now baked into how the major networks operate. For buy-side clients, the practical effect is that any consultancy without a documented compliance program is increasingly uncallable.

The legal spine has not changed. The Dirks v. SEC (1983) tipper-tippee framework still governs whether a chain of disclosure is actionable, and SEC v. Martoma remains the modern reference point for what an aggressive prosecution of expert-network channels looks like. What changed is the operational layer: networks have moved from policy documents to enforced workflow, and a new vendor stack has formed around it.

The three checkpoints, in order

Pre-call screening. Before the call is scheduled, the network verifies the expert's current employer, confirms the engagement falls outside the employer's trade window, and runs a structured questionnaire on what the expert can and cannot discuss. Dialectica's compliance module is the named example in the source; similar tooling exists internally at Guidepoint and Third Bridge. The questionnaire output is logged and retained.

In-call moderation. A trained moderator joins the call with the explicit job of cutting in if the expert drifts toward material non-public information. This was always the theoretical role of the moderator; what changed is that real-time MNPI detection is now a documented, auditable function rather than a soft norm. AI-assisted tooling (CompliAI, Smarsh) is being layered on top to flag risk language during the call, though the moderator remains the decision-maker.

Post-call transcript review. The transcript is reviewed for content that should be redacted before delivery to the client. Guidepoint and Third Bridge run automatic redaction on the post-call transcript according to the source. The redaction layer is where the compliance program leaves a paper trail that holds up in front of an investigations team.

Why the stack hardened now

Two things happened in close succession. The Capvision raid signaled that cross-border primary research carries political and regulatory risk that does not respect the buy-side's compliance comfort. Several networks paused or restructured China coverage in the months after. Then 2024 SEC enforcement actions against US-facing expert network platforms made it clear that the agency is willing to test the tipper-tippee chain through the network layer, not just at the end-investor.

The practical response from compliance teams at hedge funds and long-onlys was to tighten the gate on which networks they will engage. The question "does the consultancy have a stated compliance program" is not new. What is new is that it is now load-bearing. Networks that cannot answer it in writing are dropped from the approved-vendor list before the sourcing conversation starts.

Where the vendor stack sits

Three categories worth tracking:

• AI-assisted MNPI detection. CompliAI and Smarsh are the named vendors. The pitch is real-time risk flagging during the call, with the moderator retaining final judgment. Adoption is uneven and the technology is still being calibrated against false-positive rates that would make the moderator's job harder, not easier.
• Structured pre-call questionnaires. Dialectica's compliance module is the example. The product is less about technology than about workflow enforcement: the questionnaire is mandatory, logged, and reviewable.
• Automatic redaction. Used by Guidepoint and Third Bridge per the source. The redaction layer connects directly to the transcript pipeline, which is where transcription quality and turnaround become a compliance input rather than a convenience feature.

The last point is the one buy-side compliance teams underweight. A redaction process is only as good as the transcript it operates on. If the transcript is delayed, lossy, or inconsistently formatted, the post-call review checkpoint becomes the weakest link in the stack.

What to watch next

Three open questions for the next two to three quarters: whether SEC enforcement extends further up the chain into network operators directly; whether AI-assisted MNPI detection moves from flagging to autonomous redaction (and whether buy-side compliance teams accept that); and whether China-facing primary research rebuilds with a hardened compliance layer or stays paused.