Straiker

Primary category: agent-runtime-security. Also relevant to ai-runtime-security.

One-liner — An AI-native security platform that red-teams, protects at runtime, and discovers enterprise AI apps and autonomous agents across every layer of the AI stack.

What it does

Straiker secures the full range of enterprise AI — from AI-native apps to agents — with three modules: Ascend AI (automated attack simulation / red teaming using AI-specific safety and security threat scenarios, for risk assessment and continuous testing); Defend AI (runtime protection with real-time threat detection and automated blocking across the layers of an AI application); and Discover AI (AI agent discovery, tool monitoring, and governance visibility at scale). It uses fine-tuned models to analyze intelligence across the whole stack — user, model, application, agents, identity, and data layers — rather than filtering at the prompt level alone.

Where it sits in the stack

Spans agent-runtime-security (runtime protection + discovery of agents) and ai-runtime-security (AI-firewall-style inline detection/blocking), with a red-teaming function adjacent to ai-red-teaming. Layer: model/prompt. It works all three lethal-trifecta legs — simulates and blocks untrusted input attacks (prompt injection/jailbreaks), guards sensitive data, and constrains agent actions/egress. Combines pre-deployment testing with runtime defense in one platform.

Deployment & architecture

SaaS / API-delivered, application- and agent-layer protection (not a Kubernetes-infra agent like operant-ai). Ascend AI runs automated adversarial tests against AI apps/agents; Defend AI provides inline runtime blocking. Exact integration points (gateway, SDK, proxy) aren’t fully public — treat as medium confidence.

Positioning & differentiators

Known for being “AI-native” — purpose-built fine-tuned detection models spanning the full AI stack — and for the founder pedigree: CEO Ankur Shah led Prisma Cloud at Palo Alto Networks, and co-founder Sreenath Kurupati is also ex-Palo Alto/Intel, which shaped the unified test-plus-protect positioning. Differs from posture/governance-first noma-security/zenity/cranium by leading with runtime + red teaming, and from infra-native operant-ai by operating at the app/agent layer rather than in the Kubernetes data path. Closest neighbors: prisma-airs, hiddenlayer, straiker’s red-teaming overlaps with splxai and lakera.

Ownership, funding & M&A

Independent, VC-backed. Founded 2024; emerged from stealth March 27, 2025 with a $21M Series A from Lightspeed Venture Partners and Bain Capital Ventures. Founders: Ankur Shah (CEO) and Sreenath Kurupati, both formerly of Palo Alto Networks. HQ Sunnyvale, CA. No M&A; no seed acquisition flag. Ownership confidence: high.

CTO / hedge-fund lens

Day-2. Straiker is an early but credibly-led entrant; the combined “test before deploy + block at runtime” story is attractive if you’re shipping internal AI apps/agents and want one vendor for both. For a hedge fund, the red-teaming (Ascend AI) maps cleanly onto pre-production model/agent validation that an SR 11-7 model-risk process would want, while Defend AI covers runtime. Given its 2024 founding and single Series A, treat it as an evaluate/pilot candidate rather than an incumbent control; larger shops may prefer the breadth of prisma-airs.

Competitors / alternatives

prisma-airs, hiddenlayer, zenity, noma-security, operant-ai, splxai, lakera.

Open questions / to verify

  • Deployment/integration mechanics (gateway vs SDK vs proxy) for Defend AI.
  • Customer traction and any funding since the March 2025 Series A.
  • How its runtime blocking compares head-to-head with established AI firewalls.

Sources

History

  • [2026-06-28] Stub created from seed registry.
  • [2026-06-28] Researched; established independent/VC-backed, founded 2024, stealth exit 2025-03 with $21M Series A (Lightspeed, Bain Capital Ventures), ex-Palo Alto founders (Shah/Kurupati), Sunnyvale. Runtime + red-teaming + agent discovery scope; added ai-runtime-security as secondary category. Set ownership_confidence high.