Behavox

AI-native communications-surveillance and conduct-risk platform for regulated financial firms — ingests e-comms and voice, and uses a finance-specific LLM to flag market abuse, MNPI leakage, and misconduct.

One-liner — Behavox is a compliance-surveillance vendor that scans a firm’s e-comms and voice traffic with its own purpose-built financial-services LLM to surface market abuse, conduct risk, and insider/MNPI signals.

Categories — comms-surveillance

What it does

Behavox sits on top of a firm’s captured communications (email, chat, messaging, and transcribed voice) and runs surveillance models to generate alerts on potential market abuse (insider dealing, spoofing/manipulation), MNPI / information-barrier breaches, financial crime, conflicts of interest, anti-competitive conduct, and general misconduct. Its flagship is the Behavox Quantum surveillance platform, powered by a proprietary, finance-tuned Behavox LLM (marketed as “LLM 2.0”). The pitch is that a domain-specific model produces fewer false positives and better multilingual coverage than lexicon/keyword surveillance or generalist models, at low compute cost. Behavox Voice adds transcription across 15+ languages so voice/turret traffic can be surveilled alongside text. Behavox has also extended into regulatory archiving and, in 2025, trade surveillance (product “Polaris,” covering the major asset classes) — moving it from a pure comms-surveillance point tool toward a broader “controls platform.”

The job for a buyer: meet regulatory obligations to monitor employee communications for abuse and conduct, with materially less analyst time spent triaging noise.

Where it sits in the stack

Primary category: comms-surveillance, in the governance layer. This is a detective control — it observes and alerts on activity after the fact; it does not block prompts, redact data, or gate egress. It is not an inline prompt/egress control; it is a green-zone detective backstop that helps you discover that something bad (MNPI in a chat, collusion on a call) already happened. It complements, rather than replaces, preventive controls like DLP, information barriers, and AI runtime guardrails.

Deployment & architecture

• Deployment: SaaS (cloud) with on-premises options; SOC 2 Type 2 certified (vendor claim). SSO/SAML/Active Directory for auth.
• Data ingestion / integrations: pulls communications and content from sources such as SharePoint, Google Drive, Box, and OneDrive (per vendor). In practice comms-surveillance tools sit downstream of capture/archival systems (e.g. Smarsh, Global Relay, Microsoft Purview) — Behavox positions its own archiving as part of the ecosystem, but interoperates with existing capture estates.
• Models: proprietary Behavox LLM run in-house; vendor markets “zero third-party LLM dependencies” and “no training on user data.”

Positioning & differentiators

Behavox’s distinguishing bet is a vendor-owned, finance-specific LLM rather than rules/lexicons or wrapping a third-party foundation model — claiming better precision, multilingual voice+text alerting, and low per-day compute cost. It is known for an AI-first, “controls platform” ambition spanning surveillance, archiving, and (newer) trade surveillance.

Versus nearest neighbors:

• steeleye — combines comms surveillance with trade surveillance and record-keeping in one data layer; strong on holistic/trade-comms correlation.
• nice-actimize — incumbent, broad financial-crime and surveillance suite; deep in large banks.
• theta-lake — specialist in modern collaboration channels (Zoom, Teams, voice/video) capture + compliance.
• shield — AI-centric comms surveillance, a direct philosophical competitor on ML-driven detection.
• relativity-trace — surveillance built on the Relativity e-discovery/data platform.

Behavox overlaps most with shield and steeleye on the “AI-driven surveillance” message; its differentiator claim is the proprietary in-house LLM and voice multilingual coverage. (All capability claims here are vendor marketing and not independently benchmarked.)

Ownership, funding & M&A

• Independent, VC/PE-backed. Founded 2014; HQ London, with a large hub in Montreal and offices in New York, Belfast, Seattle, Tokyo, Singapore.
• SoftBank Vision Fund 2 invested $100M on 2020-02-24 (confirmed, vendor press release).
• HPS Investment Partners (part of BlackRock) made a $175M preferred-equity investment on 2026-06-22; Behavox concurrently repaid/retired a $70M Hercules Capital venture-debt facility (press announcement).
• Other named investors: Citigroup, Index Ventures, Hoxton Ventures. Aggregators (Crunchbase) cite ~$296M total raised — treat the aggregate as medium confidence; the two rounds above are well-sourced.
• Company states it has been profitable since 2023 and grown ~7x since 2020 (vendor claim).
• No acquisition — Behavox is the buyer’s counterparty directly; HPS/BlackRock holds a minority preferred stake, not control. ownership_confidence: high for “independent.”

CTO / hedge-fund lens

Comms surveillance is broadly Day-2 infrastructure, but for a regulated trading shop it can be effectively mandatory: firms subject to MAR (EU/UK Market Abuse Regulation), SEC/FINRA supervision, or CFTC rules are expected to monitor employee communications for market abuse and MNPI handling. A hedge fund’s actual need depends on registration and footprint:

• Large or multi-jurisdiction managers (especially UK/EU MAR-scoped, or those that have had off-channel-comms (“WhatsApp”) enforcement exposure) are the natural fit — Behavox markets relationships with several of the largest hedge-fund managers.
• Small US-only funds often satisfy obligations more cheaply via their archival vendor’s built-in surveillance (e.g. Smarsh/Global Relay) rather than a heavyweight standalone platform; Behavox is likely overkill until headcount, channels, and regulatory scope grow.

AI-prompt surveillance angle: as employees adopt genAI chat, those interactions become another communications channel a surveillance program may need to capture and review (e.g. MNPI pasted into a chatbot). Behavox’s value proposition — an LLM that reads natural-language conversations for risk — is conceptually adjacent, but I did not find an explicit, documented Behavox product feature that surveils employee genAI prompts/responses (e.g. ChatGPT/Copilot) as of 2026-06-28. Treat “AI-prompt surveillance” as a plausible roadmap adjacency, not a confirmed capability (see Open questions). For preventive controls on AI prompts/egress, that is the runtime-security/DLP layer, not Behavox.

No specific SR 11-7 / model-risk tie-in for the buyer beyond the general point that Behavox’s own surveillance LLM is itself a model a firm would need to validate/govern if relied upon for regulatory monitoring.

Competitors / alternatives

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

Open questions / to verify

• Total funding figure (~$296M) is from aggregators, not a primary filing — confirm if precision matters.
• Whether Behavox has a named, shipping feature that surveils employee generative-AI prompts/responses (vs. surveilling conventional e-comms channels) — not confirmed as of 2026-06-28.
• Exact interop with third-party capture/archival estates (Smarsh, Global Relay, Purview) — vendor lists cloud content sources but capture-vendor integrations weren’t enumerated in sources reviewed.
• Founders/CEO and the precise London-vs-Montreal HQ designation (company page labels London “HQ”; Montreal is the largest operational hub) — minor.

Sources

• Behavox Announces $100 Million Investment From SoftBank Vision Fund 2 — fetched 2026-06-28 — supports: SoftBank Vision Fund 2 $100M on 2020-02-24; comms-data platform positioning; confidence: high (primary, vendor release)
• Behavox Raises $175 Million from HPS Investment Partners, Part of BlackRock — fetched 2026-06-28 — supports: $175M preferred equity 2026-06-22, still independent, prior investors, profitable since 2023, $70M Hercules debt retired, Polaris trade surveillance; confidence: high (press announcement)
• Behavox AI / Behavox LLM product page — fetched 2026-06-28 — supports: Quantum + Behavox LLM 2.0, market-abuse/financial-crime alerting, voice+text 15 languages, SaaS/on-prem, integrations, SOC 2; confidence: medium (vendor marketing)
• Behavox — Our Company — fetched 2026-06-28 — supports: London HQ + office footprint incl. Montreal, product ecosystem, traction claims; confidence: medium (vendor marketing). Founding year 2014 / founders corroborated via Crunchbase/PitchBook search results (aggregator, medium confidence).

History

• [2026-06-28] Stub created from seed registry.
• [2026-06-28] Researched; established Behavox = independent, VC/PE-backed comms-surveillance vendor (Quantum + proprietary Behavox LLM) for MAR/SEC/FINRA market-abuse + MNPI/conduct detection across e-comms and voice. Confirmed SoftBank $100M (2020) and HPS/BlackRock $175M preferred equity (2026-06-22) — no acquisition, ownership_confidence raised to high. Set founded 2014, London HQ, SaaS/on-prem, hedge_fund_fit medium, status researched. AI-prompt/response surveillance noted as unconfirmed adjacency, not a documented feature.