AI Governance & Security Vendor Wiki

7 min read

Zscaler

Cloud-delivered security platform (SSE/SASE) that proxies user traffic through its Zero Trust Exchange; now extended to discover and govern enterprise AI use.

One-liner — Zscaler is the large public SSE/SASE vendor whose cloud proxy already sits between your users and the internet, and which now uses that position to find shadow AI, control who can use which AI apps, and block sensitive data going into prompts.

Categories — network-security-sase (primary), ai-access-governance

What it does

Zscaler runs a global cloud proxy (the “Zero Trust Exchange”) that inspects user and branch traffic instead of backhauling it to a corporate firewall. Core products are ZIA (secure web/internet access), ZPA (zero-trust private app access), and a data protection (DLP/CASB) suite. Because nearly all egress already passes through Zscaler, it can layer AI controls on top without new inline plumbing: discovering which AI apps employees use, coaching/allowing/blocking access by user or group, and applying inline DLP to prompts.

Where it sits in the stack

Foundation-layer network control under network-security-sase, with an ai-access-governance role as a secondary use of the same proxy. It primarily controls outbound data flows (governing where data can go, blocking unsanctioned AI destinations) and prevents sensitive-data leakage (inline DLP on prompts). It is the green/yellow-zone perimeter, not a model-layer guardrail.

Deployment & architecture

SaaS-delivered cloud proxy; traffic forwarded via agent (Zscaler Client Connector), GRE/IPsec tunnels, or PAC files. Inline inspection (incl. TLS) is the architecture, plus out-of-band API scanning for SaaS. AI access security features (per vendor material, 2026): shadow-AI app discovery/classification, user-based allow/block/coach policies, inline DLP with 100+ dictionaries (PII, PCI, PHI, source code), prompt content moderation, and optional browser isolation. Integrations with SIEM/SOC, IdP, and major AI providers. In Jan 2026 Zscaler announced an “AI Security Suite” (AI Asset Management, Secure Access to AI, Secure AI Infrastructure/Apps incl. red teaming). The SPLX acquisition that feeds it had closed earlier, on 2025-10-31 (see Ownership below; corrected 2026-10-23 from an earlier “announced Jan 2026” reading).

Cross-category roles (added 2026-10-23)

Survey respondents placed Zscaler in three more categories. Each is supported by Zscaler’s FY2026 Form 10-K (filed 2026-09-03):

  • dlp. Zscaler’s Data Loss Prevention “enables enterprises to alert on and/or block the transmission or sharing of sensitive data” across “web DLP, endpoint DLP, email DLP and DLP for AI systems”. It also blocks USB/print on endpoints, and “DLP policies also protect interactions with AI applications, safeguarding user prompts.” Inline DLP on the proxy is the core of it. Endpoint and email DLP are later additions. For a Zscaler shop it is often the default DLP. Classification depth, EDM and case workflow against dedicated DLP vendors were not evaluated.
  • ai-red-teaming and ai-runtime-security. “Secure AI Apps and Infrastructure … continuously red teams AI applications and models throughout the development pipeline” and “at runtime … guards AI applications and models against prompt injection, data leakage, abuse”. The red-teaming engine comes from splxai. Zscaler AI Guard (the product respondents wrote in under “AI Red-Teaming / Guardrails”) is the runtime piece. It is an API guardrail that scans prompts before the model call and responses after it, against a policy ID. Third-party AI gateways (LiteLLM, Portkey, TrueFoundry) integrate it as a guardrail provider. The same “AI Guard” brand also covers user-side AI access control (“AI Guard for Users”, see ai-access-governance), so the naming is ambiguous. Ask which AI Guard SKU is being quoted. Zscaler is a platform module in both categories, not a specialist.
  • Security operations: Red Canary MDR. Zscaler now sells a Managed Detection and Response service and an Agentic SOC product (10-K) built on the Red Canary acquisition. This makes Zscaler an MDR provider. Cross-listed in ai-soc-analysts on 2026-10-24, when MDR providers were placed in that category (decision in log.md). The Agentic SOC product itself has not yet been benchmarked against the AI-SOC startups.

Positioning & differentiators

Known for scale and an installed inline footprint: if a shop already routes egress through Zscaler, turning on AI discovery/DLP is incremental rather than a new deployment. Differs from API-first CASB/DSPM tools by being inline-first. Nearest neighbors: netskope (closest SSE/SASE competitor, also pushing AI controls), palo-alto-networks (Prisma Access / Prisma AIRS), cisco, cato-networks, forcepoint. Versus dedicated AI-runtime/guardrail vendors, Zscaler’s AI governance is a perimeter overlay, not deep model-prompt inspection.

Ownership, funding & M&A

Public company, NASDAQ: ZS (IPO March 2018); HQ San Jose, California; founded 2007 by Jay Chaudhry. Confirmed public status via vendor/IR press releases (as of 2026-06-28). No ownership ambiguity. Acquisition ledger (Zscaler FY2026 Form 10-K, Note 6, filed 2026-09-03):

  • Red Canary (MDR): definitive agreement announced 2025-05-27, closed 2025-08-01 for $651.4M cash plus $20.2M restricted stock. Trade press reported ~$675M at signing, which matches the headline figure before adjustments. Red Canary initially runs as a separate business unit (“Red Canary, a Zscaler company”). It is the Red Canary that survey respondents wrote in under SIEM.
  • splxai (AI red teaming): closed 2025-10-31, $40.6M cash plus $16.6M restricted stock, publicly announced 2025-11-03.
  • SquareX (browser security): closed 2026-02-05, $112.8M cash plus $37.4M restricted stock.
  • symmetry-systems (DSPM): the 10-K states Zscaler “acquired all outstanding equity” on 2026-05-27 for $152.9M cash plus $16.9M restricted stock. So the deal has closed. The symmetry-systems page (last verified 2026-07-27) still shows it as pending and needs updating.

CTO / hedge-fund lens

Day-1 if you want a single cloud-delivered egress/secure-access layer for a distributed or remote workforce; the AI-governance features are then a low-friction Day-1.5 add-on for controlling ChatGPT-style shadow AI and stopping data leakage into prompts. For a small fund this is a heavyweight enterprise platform — capable but priced and scoped for larger estates; a 50-person shop may get most of the AI-access value from a lighter SSE or a browser/DLP point tool. No direct SR 11-7 / model-risk function; it governs access to AI, not model validation.

Competitors / alternatives

netskope, palo-alto-networks, cisco, cato-networks, forcepoint.

Open questions / to verify

  • Independent (non-marketing) validation of AI-access DLP efficacy and prompt inspection depth.
  • How SPLX (red teaming) and AI Guard (runtime) are packaged and priced together, and which “AI Guard” SKU covers what.
  • Whether Red Canary MDR stays a separate business unit or becomes “Zscaler MDR”. Are non-Zscaler customers still supported?
  • Depth of Zscaler endpoint and email DLP vs. its inline proxy DLP.
  • Exact pricing/packaging of the AI Security Suite vs base ZIA/data-protection licenses.

Sources

History

  • [2026-06-28] Stub created from seed registry.
  • [2026-06-28] Researched; confirmed public (NASDAQ: ZS, IPO 2018), HQ San Jose, founded 2007; documented AI access governance / shadow-AI / inline-DLP positioning; set ownership: public (high). One vendor source cached (marketing).
  • [2026-10-23] Cross-listed in dlp, ai-red-teaming and ai-runtime-security from Zscaler’s FY2026 10-K product descriptions plus AI Guard integration docs, after Oct-2026 CTO survey write-ins (“Zscaler” under DLP; “Zscaler AI Guard” under AI Red-Teaming/Guardrails). Verified Red Canary acquisition (announced 2025-05-27, closed 2025-08-01, $651.4M cash). Corrected the SPLX timing (closed 2025-10-31, not announced Jan 2026). Added SquareX (closed 2026-02-05) and found that Symmetry Systems closed 2026-05-27 per the 10-K, so the symmetry-systems page is stale. Set market_position labels. Not cross-listed to an MDR/SOC category (none exists for MDR, and Agentic SOC was not researched).
  • [2026-10-24] Cross-listed in ai-soc-analysts (Red Canary MDR + Agentic SOC) after the decision to place MDR providers in that category; a survey respondent wrote Red Canary in under SIEM/Logging (in production).

Graph View

Backlinks