AI Governance & Security Vendor Wiki

6 min read

Aurascape

Researched 2026-06-28. Primary category: ai-access-governance; also ai-runtime-security.

One-liner — An AI-native security platform that gives enterprises visibility into and control over every employee/app interaction with AI tools (sanctioned and shadow), with prevention-first data protection across text, code, image, audio, and video.

What it does

Aurascape watches how people and applications in an organization actually use AI — public chatbots, embedded AI features, copilots, and agentic tools — and enforces policy on those interactions in real time. It decodes prompt/response traffic across what it markets as 1,500+ AI applications, classifies the content (including non-text modalities), and can block or remediate risky actions such as sensitive data being pasted into a public model. The core jobs: discover shadow AI, prevent data leakage to AI services, and apply granular guardrails without simply blocking all AI use. Marketing frames this as “AI Activity Control” and “prevention-first.”

Where it sits in the stack

  • Primary: ai-access-governance — it governs who can use which AI tools and what data may flow to them, the classic shadow-AI discovery + access-control job. Layer: model-prompt.
  • Secondary: ai-runtime-security — because it inspects and can block prompt/response traffic inline, it overlaps with the AI-firewall/guardrail category.
  • Risk role: mainly preventing sensitive-data exposure and controlling outbound data flows (stopping confidential data from leaving to external AI services); to a lesser degree screening untrusted content in responses. It is fundamentally an egress/DLP-for-AI control sitting at the boundary between the corporate (green/yellow) zone and external AI services (red zone).

Deployment & architecture

Hybrid, per the lead investor’s description:

  • Cloud-native inline proxy for network-level monitoring of AI traffic.
  • Endpoint agent for device-level visibility (catches browser plugins, local tools, traffic that bypasses the network path).
  • Coverage spans LLMs, embedded AI, and shadow-AI tools; multimodal classification (text, code, image, audio, video).
  • Integrations noted: Microsoft Copilot monitoring. Broader SIEM/IdP/DSPM integration not yet confirmed from primary sources.

This proxy-plus-agent model is closer to a SASE/SSE data-protection architecture than a pure API-based guardrail — unsurprising given the founders’ Zscaler/Palo Alto/Netskope background.

Positioning & differentiators

Aurascape positions as AI-native and prevention-first, arguing that legacy firewalls, regex DLP, and SASE tools can’t parse the probabilistic, multimodal nature of AI traffic. Claimed differentiators (vendor/investor, treat as marketing): breadth of app coverage (1,500+), multimodal detection, low false positives, and 14 patents filed (2 approved as of the Menlo post).

Versus neighbors:

  • vs witnessai — both do AI access governance/observability with inline enforcement; close competitor.
  • vs harmonic-security and wald-ai — overlap on preventing sensitive-data leakage to GenAI; Harmonic leans browser/data-centric.
  • vs prompt-security — Prompt leans toward runtime prompt-injection/guardrail firewalling; Aurascape leans access-governance + DLP.
  • vs lanai — both target shadow-AI discovery; Lanai is more endpoint/edge-discovery focused.
  • vs cyberhaven — Cyberhaven is data-lineage DLP extending into AI; Aurascape is AI-first.
  • vs nudge-security and reco — those lean SaaS-posture/identity discovery; Aurascape is deeper on inline AI-traffic control.

Ownership, funding & M&A

  • Independent, VC-backed. No acquisition found; confidence high that it is independent as of 2026-06.
  • Founded 2024, Santa Clara, CA. Launched from ~1 year of stealth on 2025-04-08.
  • Seed: $12.8M, August 2024, led by Mayfield Fund (with Celesta Capital, StepStone Group, AISpace, and Mark McLaughlin participating) — per aggregator/press, medium confidence.
  • Series A: $50M, April 2025, co-led by Mayfield Fund and Menlo Ventures; strategic investors include former Palo Alto Networks CEO Mark McLaughlin, former Symantec CEO Greg Clark, Walden International’s Lip-Bu Tan, and former Zscaler CSO Manoj Apte.
  • Reported total funding: ~$62.8M. Note: the launch PR headlines “$50M” (the Series A); the $62.8M figure is the seed + Series A total per press/aggregators.
  • Founders: Moinul Khan (CEO; ex-Zscaler SSE/data-protection lead, ex-Palo Alto Networks, ex-Netskope), Patrick Xu (CTO). Additional co-founders (Viswesh, Liang Li) per aggregators — lower confidence on exact roles. Team drawn from Palo Alto Networks, Zscaler, Netskope, Google, Amazon.

CTO / hedge-fund lens

This is a Day-1-relevant control for any firm where employees are already using AI tools (i.e., everyone): it answers “what AI are my people using and what data is going to it?” For a hedge fund, the strongest fit is the egress/DLP angle — preventing MNPI, positions, or client data from leaking into public models — plus an audit trail of AI usage that supports governance and exam readiness. SR 11-7 model-risk relevance is indirect: Aurascape governs usage and data flow, not model validation, so it complements rather than satisfies model-risk programs.

Caveats for a smaller shop: it’s an early-stage (2024) company with a proxy + endpoint-agent footprint, which is heavier to deploy than a pure-API guardrail and competes in a crowded field. Reference customers, pricing, and SOC 2/compliance posture are not yet established here and should be diligenced. Fit rated medium pending that.

Competitors / alternatives

witnessai · harmonic-security · wald-ai · prompt-security · lanai · nudge-security · reco · cyberhaven

Open questions / to verify

  • Exact total funding: reconcile “$50M” (PR headline) vs “$62.8M” (reported total). Seed round amount/date needs a primary source.
  • Full co-founder roster and roles (Patrick Xu / Viswesh / Liang Li currently aggregator-sourced).
  • Integration depth: SIEM/SOAR, IdP/SSO, DSPM, MCP support — not confirmed from primary sources.
  • Compliance posture (SOC 2, ISO), pricing model, and named reference customers.
  • Real-world false-positive rate and deployment friction of the inline-proxy + agent model.

Sources

History

  • [2026-06-28] Stub created from seed registry.
  • [2026-06-28] Researched; established founded 2024 (Santa Clara, CA), launched from stealth Apr 2025, $50M Series A (Mayfield + Menlo) + $12.8M seed (~$62.8M reported total), independent (no M&A) at high confidence, CEO Moinul Khan (ex-Zscaler/PAN/Netskope). Added ai-runtime-security as secondary category given inline prompt/response enforcement. 4 sources cached.

Graph View

Backlinks