Hadrius

One-liner — An AI-native compliance platform for SEC/FINRA-regulated US firms that bundles comms supervision, marketing review, personal-trading oversight and testing into one system of record — and is one of the few vendors that archives employees’ ChatGPT and Claude conversations as a supervised books-and-records channel.

What it does

Hadrius sells “agentic compliance infrastructure” to registered advisers and broker-dealers, pitched as a replacement for the stack of legacy point tools a compliance team otherwise runs. Six modules (vendor’s framing, 2026-08-18):

• Communications — supervision of email, chat, social media and AI-assistant conversations, with 17(a)-4 / WORM retention and what the vendor calls “zero-data-retention AI” review.
• Marketing — AI-assisted review and approval workflow for advertising and marketing material (SEC Marketing Rule territory).
• People Oversight — personal trading, attestations, certifications.
• Account Surveillance — employee and firm trades checked against restricted lists and blackout windows.
• Testing Program — automated compliance testing with audit-ready reporting.
• Compliance-manual implementation and branch-location oversight.

Multi-tenant workspaces let compliance consultants run many client firms from one console — a real channel in the RIA market, where much of the compliance function is outsourced.

The part that matters for an AI stack

On 2026-07-28 Hadrius added Claude to its communications archive, built on Anthropic’s Claude Compliance API; ChatGPT capture was already supported. Firms on Claude Enterprise can archive employees’ AI conversations alongside email and chat, with the same surveillance applied. Capture is agentless — nothing installed on employee devices, no change to how employees use the assistant — and runs on Hadrius infrastructure. Hadrius is named on Anthropic’s own Compliance API partner list, so this is corroborated, not just vendor copy.

This closes the gap flagged elsewhere in this wiki: most comms-surveillance incumbents apply AI to the archive (LLM alert triage) but do not treat the AI assistant itself as a channel. The regulatory logic is simple and hard to argue with — SEC recordkeeping attaches to any channel where employees conduct business, and an adviser working through a client question with Claude is conducting business.

Where it sits in the stack

• Primary category: comms-surveillance (governance layer); also relevant to enterprise-grc.
• Detective control, not preventive. It records and reviews after the fact; it does not block a prompt or an egress. If you want an employee stopped from pasting the position book into a chatbot, that is dlp / ai-access-governance, not this.
• Scope-boundary note: the category page draws a line between capture/archival (Smarsh, Global Relay, Purview) and surveillance/analysis. Hadrius deliberately does both — it is the archive and the supervision layer. That is the pitch for small firms and a limitation for firms with an entrenched archive they will not move.

Positioning & differentiators

• vs behavox, shield, nice-actimize — those are enterprise conduct/market-abuse surveillance engines aimed at banks and large asset managers, usually layered on a separate archive. Hadrius is aimed down market at the US registered-adviser/BD compliance program as a whole, and covers non-surveillance chores (attestations, marketing review, testing) those tools do not.
• vs steeleye — SteelEye couples comms with trade surveillance under MAR for EU/UK-facing firms. Hadrius is US-rule-shaped (SEC/FINRA, 17a-4, Marketing Rule) and its trading angle is personal/employee account surveillance, not market- abuse modeling on the firm’s order flow.
• vs greenboard — the closest comparable: same YC lineage, same buyer, overlapping module map. The clearest split today is AI-assistant capture, which Hadrius ships and Greenboard does not list. Hadrius is also roughly 3x the headcount (80 vs 26 per YC, 2026-08-18) and better capitalized ($27M vs $20M).
• vs theta-lake — Theta Lake specializes in modern collaboration (video, voice, UC) capture and risk detection; deeper on those channels, far narrower as a compliance program.

Who should choose them / anti-fit

Good fit: a US RIA, broker-dealer or private fund — anywhere from a few people to a few thousand — that wants one platform for the whole compliance program instead of an archive plus a surveillance tool plus a marketing-review inbox plus a spreadsheet of attestations. Sharpest fit if you have deployed an enterprise AI assistant and need a recordkeeping answer for it, which is a live and under-served problem.

Anti-fit: EU/UK firms whose obligation is MAR-shaped; firms needing genuine trade/market-abuse surveillance on firm order flow; tier-1 institutions with an incumbent Smarsh/Global Relay archive and a NICE Actimize-class surveillance stack they are not replacing; anyone requiring self-hosted or in-tenant deployment.

Known weaknesses / gotchas

• Young and unproven at scale. Founded 2023. Impressive customer count, but the reference base for hedge-fund-scale complexity is thin; ask for references at your size and strategy.
• Efficacy numbers are marketing. “99% reduction in false positives,” “70% less manual review,” “20 hours a week saved” are all vendor/customer-reported with no defined denominator. The $3T AUM figure is a sum across the client base, not a measure of depth.
• Consolidation cuts both ways. One platform for archive + supervision + testing means one vendor’s outage, one vendor’s retention posture, and a painful exit if you later need to migrate a 17a-4 archive out.
• AI-assistant capture requires Claude Enterprise. The Compliance API is not available on lower Claude tiers, so the marquee capability has a licence prerequisite on the model-vendor side.
• Certifications are vendor-stated (ISO 27001, SOC) and were not independently verified here — ask for the reports.

Deployment & data handling

Multi-tenant SaaS; data leaves your tenancy by design — that is what an archive is. Capture for AI assistants is agentless and API-based (Claude Compliance API), running on Hadrius infrastructure. The vendor markets “zero-data-retention AI” for the review layer, which should mean model providers do not retain content passed for analysis — get this in writing, distinguishing (a) retention by the LLM provider, (b) retention by Hadrius, and (c) any use of customer content for model training. None of the three is spelled out in public sources.

Integrations & partnerships

Email, chat, SMS, social media; ChatGPT and Claude Enterprise conversation capture via Anthropic’s Compliance API. Named customers include BBVA, Altruist, Allworth Financial, World Investment Advisors, M1 Finance (vendor). MCP support: unverified.

Compliance & FS traction

Purpose-built for SEC/FINRA obligations: 17(a)-4/WORM recordkeeping, supervision, Marketing Rule review, personal-trading and code-of-ethics oversight, exam-ready evidence. Vendor claims 500+ financial institutions and compliance managed across “$3T+ AUM”; fs_traction: claimed — named logos are real, the aggregate stats are self-reported.

Commercial

Not disclosed. No public pricing; expect per-seat or per-firm annual subscription with module-based tiering, as is standard in this segment.

Competitors / alternatives

greenboard · behavox · steeleye · shield · theta-lake · nice-actimize · relativity-trace

Open questions / to verify

• Data handling specifics — retention periods, sub-processor list, whether customer content (including archived AI conversations) is ever used for model training, and what “zero-data-retention AI” contractually means.
• Which ChatGPT tier the existing OpenAI capture requires, and by what mechanism (compliance API vs admin export) — the Claude path is documented, the ChatGPT one is a bare vendor claim.
• Whether Copilot / Gemini / other assistants are on the roadmap as capture channels.
• Independent confirmation of ISO 27001 and SOC 2 reports.
• Pricing bands, and whether the consultant multi-tenant edition is priced differently.
• Post-Series A cap table and board composition.

Sources

• Hadrius — Announcing our $22 Million Series A — fetched 2026-08-18 — supports: $27M seed+A / $22M Series A, CRV lead, YC + Pathlight, 2026-07-14, founders, NYC HQ, 500+ firms, named customers; confidence: high for funding, low for efficacy stats (vendor).
• Hadrius — homepage / product modules — fetched 2026-08-18 — supports: six-module map, 17(a)-4/WORM, zero-data-retention AI claim, consultant workspaces, vendor-stated ISO 27001 / SOC / GDPR; confidence: med (vendor marketing).
• Hadrius Adds Claude to Its Communications Archive — PRNewswire — fetched 2026-08-18 — supports: 2026-07-28 Claude integration, agentless capture, Claude Enterprise requirement, prior ChatGPT support, Stewart quote; confidence: high (vendor PR, corroborated below).
• Anthropic — Claude Compliance API integrations — fetched 2026-08-18 — supports: Hadrius named as a Compliance API partner; Enterprise vs Platform access tiers; independent corroboration of the integration; confidence: high (primary, the API’s own vendor).
• Hadrius — Y Combinator company profile — fetched 2026-08-18 — supports: YC W23, founded 2023, NYC, team size 80, founder backgrounds, $3T AUM claim; confidence: high for identity facts, low for the AUM claim.
• Cached: raw/sources/2026-08-18--hadrius--funding-and-yc-profile.md, raw/sources/2026-08-18--hadrius--claude-compliance-api-archive.md.

History

• [2026-08-18] Page created and researched in one pass (survey-v2.csv row → wiki-create + wiki-research). Established: founded 2023, NYC, YC W23, independent, $27M raised ($22M Series A led by CRV, 2026-07-14), six-module SEC/FINRA compliance platform. Key finding: archives ChatGPT and Claude Enterprise conversations as a supervised recordkeeping channel via Anthropic’s Compliance API (2026-07-28), corroborated by Anthropic’s partner list — the AI-channel-capture gap flagged on steeleye. Efficacy and AUM stats recorded as vendor claims, not facts.