The new compliance stack for primary research

The primary-research compliance stack used by expert networks looked very different three years ago. Then came the Capvision raids in May 2023 and a sequence of SEC enforcement actions against US platforms through 2024. Every major network has since rebuilt the workflow around three checkpoints and a layer of new tooling.

The pre-call gate is the first one. Experts are screened on current employer, employment status, and trading-window posture before a call is scheduled. Dialectica has productized this as a compliance module that surfaces conflicts before the moderator joins. Guidepoint, GLG, and Third Bridge run their own variants of the same questionnaire flow, with the screening output logged as the audit trail.

The second checkpoint is live moderation. A trained moderator sits on every call, listens for material non-public information in real time, and intervenes when an expert drifts toward disclosing it. This is the most expert-network-specific layer of the stack and the one most resistant to automation, because the judgment call (is this material? is it non-public?) requires industry context the model does not have.

The third is post-call review. Transcripts are scanned for MNPI before delivery to the client. Guidepoint and Third Bridge use automated redaction; vendors including Smarsh and CompliAI sell AI-assisted MNPI detection that flags candidate passages for human review. The reviewer is still a person. The model narrows the search.

What the law actually says

The legal anchor for all of this is Dirks v. SEC, the 1983 Supreme Court decision that established the tipper-tippee framework for insider trading. A tippee inherits the tipper's duty if the tipper breached a fiduciary duty in passing the information and the tippee knew or should have known. The doctrine got sharpened in SEC v. Martoma, the SAC Capital case that produced a nine-year sentence in 2014 and remains the case compliance teams cite when training moderators.

The practical effect on expert networks: a consultancy that cannot point to a documented compliance program is now treated by buy-side compliance teams as a chain-of-custody risk. The screening, moderation, and review steps are not just protective for the platform. They are the artifact the hedge fund's compliance team needs to greenlight the engagement at all.

Where the China overlay sits

The Capvision raid added a second axis. Several US expert networks paused or restructured their China call programs after the coordinated raids on consulting firms in 2023 that swept up Capvision, Bain, and Mintz Group. The compliance question shifted from MNPI under US securities law to the broader question of whether a China-sourced expert call could create exposure under China's anti-espionage and data-security regimes.

The answer most platforms landed on: tighter expert vetting in mainland China, a narrower roster, and explicit topic restrictions that screen out anything brushing state-owned enterprises, semiconductor policy, or military adjacency. The compliance overhead per call rose. The number of calls per quarter fell.

The tooling layer

The new tools cluster in three places:

1. Pre-call: structured questionnaires that capture employer, role, trading-window posture, and topic scope. Dialectica's compliance module is the most-cited example.
2. In-call: not much, yet. Real-time MNPI detection is hard because the audio is noisy, the context is industry-specific, and the cost of a false negative is large. Most platforms still rely on the moderator.
3. Post-call: automated redaction and AI-assisted MNPI flagging at the transcript layer. CompliAI and Smarsh are the names that come up most. Guidepoint and Third Bridge run automated redaction in production.

The transcript layer is where the compliance stack and the data-vendor stack meet. A clean, accurate, timestamped transcript is the prerequisite for both: the compliance team uses it to defend the call, and the analyst uses it to source the memo. Sloppy transcription is now a compliance problem, not just a research problem.

What to watch through 2026: whether the SEC opens a second front on the platforms themselves rather than the trading firms, and whether China's compliance overlay tightens further around primary research as a category. Both would push the stack toward more checkpoints, not fewer.