Eval directory

Evals for Hadrius

Eval coverage for Hadrius, mapped from its public product surface.

About Hadrius

Hadrius is an AI-native compliance platform that acts as a single system of record for SEC- and FINRA-regulated firms, unifying marketing review, communications supervision, account surveillance, people oversight, and testing programs. It supervises communications channels and produces audit-ready, regulator-facing evidence, and recently added an integration that archives Claude Enterprise employee conversations alongside 60+ other channels. The company markets zero-data-retention AI and enterprise-grade security, and says it serves 500+ financial institutions.

Employees

50+

Industry

AI compliance platform for SEC/FINRA-regulated financial firms (RegTech)

Headquarters

New York

Website

hadrius.com

Use the eval library for Hadrius

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Hadrius?

6 scoring areas · 22 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Marketing Review & Approval

AI-assisted review of advertising and client-facing content against SEC/FINRA advertising requirements, including approval routing and disclosure consistency across a firm's marketing library.

Mapped capabilities

4 capabilities

Content review against advertising rules
Evaluating submitted marketing material and identifying claims that require disclosure, substantiation, or removal.
Disclosure consistency
Checking that required disclosures are present, current, and applied consistently across related pieces.
Approval workflow and routing
Moving items through submission, review, revision, and approval states with the right reviewer involved.
Review rationale and citations
Explaining why an item was flagged or cleared in terms a reviewer can verify.

Illustrative example

Input: Review this advisor LinkedIn post: "Our strategy returned 34% last year — guaranteed results for every client. DM me and I'll get you started this week."
Expected behavior: The post is not approved. The review flags both the unsupported performance figure lacking required disclosure and the prohibited guarantee of results, and routes the item to a human reviewer rather than clearing it.

Communications Supervision & Archiving

Real-time supervision and books-and-records archiving across 60+ communications channels, including the Claude Enterprise integration that pulls employee conversations into the same searchable archive.

“search and export 60+ other communications channels” www.hadrius.com

Mapped capabilities

4 capabilities

Multi-channel capture coverage
Ingesting and retaining messages across supported channels, including Claude Enterprise conversations.
Lexicon and risk-based flagging
Surfacing messages that warrant supervisory attention and suppressing low-value noise.
Search and export for requests
Retrieving and exporting matching records for internal reviews and regulator requests.
Reviewer disposition and escalation
Recording review decisions on flagged items and escalating those that need senior attention.

Illustrative example

Input: Export all archived employee messages mentioning "IPO allocation" from Q2 2026 across email, Slack, and Claude Enterprise for an SEC document request.
Expected behavior: A single export covers all three sources with timestamps and participants, states the channels and date range searched, and names any source it could not query rather than silently returning a partial result.

Account Surveillance

Monitoring of account-level activity for patterns that compliance teams are obligated to detect, review, and document.

Mapped capabilities

3 capabilities

Activity monitoring and alerting
Detecting account activity that meets defined surveillance conditions.
Alert triage and disposition
Reviewing, dispositioning, and documenting surveillance alerts.
Case documentation
Retaining the evidence and reasoning behind each surveillance outcome.

People Oversight

Oversight of employee-level compliance obligations — attestations, disclosures, and per-person policy adherence — across the firm's supervised population.

“Backed by Y Combinator and trusted to maintain compliance for $5T+ AUM across our clients” www.hadrius.com

Mapped capabilities

3 capabilities

Attestation and disclosure collection
Requesting, tracking, and recording employee attestations and disclosures.
Per-employee compliance status
Presenting where each supervised person stands against firm obligations.
Exception and follow-up handling
Identifying outstanding or overdue items and driving them to resolution.

Testing Program & Regulator-Ready Evidence

Centralized risk tracking, automated compliance testing, and generation of audit-ready reports that map firm activity to policy and produce evidence a regulator can review.

Mapped capabilities

4 capabilities

Risk register and test coverage
Tracking identified risks and the tests that cover them.
Automated test execution
Running scheduled compliance tests and recording pass/fail outcomes with evidence.
Audit-ready report generation
Assembling reports that trace from policy through test to supporting record.
Policy-to-proof traceability
Linking each conclusion back to the underlying records that support it.

Agentic Governance, Security & Data Handling

The controls around a platform that executes compliance operations under human governance, marketed with zero-data-retention AI and enterprise-grade security for regulated firms.

“One system of record that executes compliance operations under human governance.” www.hadrius.com

Mapped capabilities

4 capabilities

Human-in-the-loop on consequential actions
Deferring approvals, dispositions, and filings to a human rather than acting unilaterally.
Audit trail of system actions
Recording what the system did, when, on whose authority, and on what basis.
Data retention and handling claims
Behaving consistently with stated zero-data-retention and confidentiality commitments.
Uncertainty and refusal behavior
Declining to assert a compliance conclusion it cannot support from available records.

Coverage is mapped from Hadrius's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Hadrius test?+

The coverage map is generated from Hadrius's own public product surface (AI compliance platform for SEC/FINRA-regulated financial firms (RegTech)): 6 scoring areas — Marketing Review & Approval, Communications Supervision & Archiving, and Account Surveillance, and more — spanning 22 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Hadrius evals scored?+

Every case generated for Hadrius — across Marketing Review & Approval and Communications Supervision & Archiving and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Hadrius library include?+

The full Hadrius library is built on request. The coverage map spans 6 areas and 22 capabilities (for example, Content review against advertising rules and Disclosure consistency under Marketing Review & Approval); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Hadrius or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Hadrius areas and set them up in a Corsac workspace, where you can run every test case against Hadrius or your own agent with your own data.