Behavox: Initiation of Coverage

GAII Intelligence defines the Risk, Compliance & Surveillance market as software that applies AI to the supervisory obligations of investment firms, banks and trading venues: capturing and retaining communications and trading records, detecting misconduct and market abuse, reviewing regulated content, and evidencing those controls to regulators. Buyers are chief compliance officers, heads of surveillance and conduct risk, and the technology and records teams that support them.

Capabilities we treat as mandatory

• Complete, defensible capture: records of every in-scope channel or trading system, reconciled and retained to regulatory standards such as SEC 17a-4.
• Explainable detection: every alert or finding traceable to the policy, rule or signal that produced it, in a form a regulator can examine.
• Institutional data handling: a stated position on where client data goes, how long it is held, and whether it is used for model training.

Capabilities we treat as differentiating

• False-positive reduction: measurable cuts in alert volume without loss of true positives, with a disclosed method.
• Agentic triage: agents that disposition first-level alerts and assemble cases under human supervision.
• Holistic coverage: communications, trade and conduct data analysed together rather than in separate tools.
• AI-channel governance: capture and supervision of employees' use of AI assistants as a regulated channel.

Behavox competes with Smarsh, Global Relay and NICE Actimize in communications surveillance and archiving, with Theta Lake and Shield in AI-led supervision, and — once Polaris ships — with Nasdaq SMARTS, NICE Actimize, Eventus and TradingHub in trade surveillance.

The following are the capabilities and characteristics that, in our assessment, most distinguish Behavox within the Risk, Compliance & Surveillance market. Each is drawn from publicly documented product capability, published interviews with management, or independent review.

• Control breadth: Regulatory change, information barriers, conflicts, communications, voice and insider surveillance, archive and agents from one vendor.
• Tier-one references: BNY enterprise rollout across 16+ languages and Mizuho Securities live within three months; ten of 24 G-SIBs as customers (company).
• Domain models: Risk policies trained on more than a million labelled examples, with ranked, plain-language alert explanations.
• Financial position: Profitable in 2024 with audited revenue of $35.6m; $175m from HPS; venture debt repaid.
• Assurance: Several years of clean SOC 2 Type II reports from Schellman and ISO 42001 certification for AI management.
• Agents alongside incumbents: Digital Employees triage first-level alerts in Smarsh, Global Relay, NICE, Nasdaq and Eventus queues.

The following are factors that prospective clients should weigh, and areas where our public review did not find sufficient evidence to reach a settled view. Several are common across vendors at this stage of the market and are not specific to Behavox. We encourage prospective clients to raise each of them directly with the company; we have done the same, and will update this note as answers arrive.

• Trade surveillance timing: Polaris was targeted for late 2025, then the second quarter and then summer 2026; no general-availability announcement was located.
• Integrations: No public API (MCP and REST "coming soon"), and OMS/EMS and HR connectors are not published.
• Assurance gaps: ISO 27001 appears on one product page but not the trust centre; SSO and SCIM were not documented.
• Model disclosure: Detection models are proprietary, but the frontier models behind agents are not named; buyers should request providers and terms.
• Evidence: Recall and false-positive reductions are vendor-run; profitability statements conflict with the audited FY2023 loss.
• Scale: Third-party data show headcount down from about 300 to 235 since 2023.
• Ownership terms: HPS holds preferred equity on undisclosed terms; SoftBank sold part of its stake.

Behavox Ltd. was incorporated in London in 2014 by Erkin Adylov, formerly a Goldman Sachs equity researcher and Man Group portfolio manager. It has offices in London, New York, Montreal, Seattle, Singapore, Tokyo and Milan, acquired Mosaic Smart Data in 2024 and Apiax, a Zurich compliance-rules company, in October 2026.

Behavox raised early capital from Citi, Index Ventures and Hoxton, $100m from SoftBank Vision Fund 2 in 2020 and $175m of preferred equity from HPS Investment Partners in June 2026, when BlackRock became a significant shareholder and SoftBank sold part of its stake. Audited FY2024 revenue was $35.6m with a small profit; FY2025 ARR grew more than 30% (company).

Company-reported figures are shown as such and are not independently verified. Verification is carried into the vendor questionnaire.

Behavox organises its products around preventive, detective and corrective controls. Figure 2 sets out our assessment of how it maps onto the compliance and surveillance workflow.

Quantum

Communications and voice surveillance with AI risk policies in 50+ languages.

Poseidon, Trident and Pathfinder

Information barriers, conflicts and personal dealing, and regulatory change.

Intelligent Archive

WORM archive with legal hold, Relativity export and free data exports (2024).

Agents and Polaris

Digital Employees and Scout for triage and investigation; Polaris trade surveillance in preview.

Scope of the product today

We did not locate a public API, published OMS/EMS connectors or general availability of Polaris.

Behavox trains its own detection models on more than a million labelled compliance examples and runs them on GPUs in single-tenant Google Cloud environments; a 2025 migration cut alerts by more than 60% (company). Its agents use frontier models through zero-retention endpoints.

Explanations rank the signals behind each alert against a risk taxonomy, and models are validated to bank model-risk guidance. Evaluation figures are vendor-run.

Behavox captures 150+ text channels, voice from turrets and recorded lines, and interactions with AI assistants, with lineage from ingestion to alert and an archive claiming SEC 17a-4 and MiFID II compliance.

Its detection library combines AI risk policies, 100+ scenarios and specialist classifiers; trade data enter through Polaris and Mosaic.

Behavox fits the surveillance team's daily workflow: fewer, better-explained alerts, first-level triage by agents, and case assembly across chats, voice and archive.

Integration with order-management and HR systems is the weakest link; there is no public API.

Behavox holds SOC 2 Type II reports from Schellman and ISO 42001, runs single-tenant on Google Cloud with regional residency and a UK sovereign option, and publishes a trust centre. Figure 3 sets out what our public review located.

We would encourage buyers to request ISO 27001 status, SSO and SCIM, model providers and a no-training clause.

The offsetting factors are references and finances: tier-one banks in production and a profitable, debt-free balance sheet.

Behavox does not publish pricing. It claims lower export costs, consolidation of several archive vendors and agents priced at a fraction of staff cost.

Buyers should seek quantified references and contract terms covering Polaris delivery.

Behavox sits in the Emerging quadrant of the segment matrix, just below the investment-utility boundary, with the broadest product range among the five companies we cover.

GAII Risk, Compliance & Surveillance coverage, 20 October 2026, scored on the segment variant of Data & Content and Workflow Fit; not directly comparable with Research & Intelligence scores.

11.1 Where Behavox is strongest relative to peers

• Against archive incumbents, on AI-native detection and agent triage.
• Against point solutions, on breadth of controls.
• Against peers, on multilingual voice and text models.

11.2 Areas of narrower functional focus

Behavox has built breadth and models ahead of integrations and trade surveillance.

• Trade: Polaris not yet generally available.
• Integrations: no public API.
• Assurance: ISO 27001 and SSO unverified.
• Evidence: vendor-run metrics.

Scores are on a 0–10 scale. Each dimension score is the mean of its sub-criteria, and the GAII Overall Score is the weighted mean of the six dimensions. The detailed sub-criteria scores that build each dimension are set out in Appendix A.

These are the areas where our public review did not find sufficient evidence to form a settled view. We have put them to the company and will update this note when they are answered. Prospective clients may find them a useful agenda for their own evaluation call.

• Polaris general-availability date and current clients.
• Public API and MCP timing.
• ISO 27001 status.
• SSO/SAML and SCIM.
• Frontier model providers and no-training terms.
• Whether client labels train shared models.
• FY2025 revenue.
• HPS preferred-equity terms.
• Third-party attestation of 17a-4 compliance.
• Pricing.
• Quantified client references.
• Headcount and support capacity.

This note is an assessment of a vendor against a defined framework, not a recommendation to buy or to decline. The following are the considerations we would put at the centre of an evaluation of Behavox specifically.

If you run compliance or surveillance

• Pilot Quantum against your current lexicon.
• Test agent triage on existing queues.
• Confirm Polaris timing before replacing trade surveillance.

If you are responsible for technology or vendor risk

• Open with Figure 3; request ISO 27001 and SSO status.
• Plan integrations without a public API.
• Review model providers.

If you are assessing the category rather than the company

• AI is moving surveillance from lexicons to policies.
• Agents now triage across vendors.
• Suites are converging communications and trade surveillance.

Each dimension score in Section 12 is the mean of the sub-criteria below. Data & Content and Investment Workflow Fit use the Risk, Compliance & Surveillance segment variant.