About Dash Security
Dash Security is building the security layer for the agentic enterprise - so teams can let AI run, securely. Enterprises are adopting AI agents across coding tools, SaaS, cloud, and endpoints, and traditional security stacks don’t work where agents interpret instructions, plan, and act at runtime. Dash discovers every agent, MCP server, skill, and interaction; captures full agentic sessions end to end; and observes, governs, and acts across every agent and interaction, everywhere they run - instantly. We exist to close the agent security gap with intent-aware runtime protection that scales with enterprise AI adoption.
Role Overview
We are seeking an AI Security Researcher to join Dash’s core research team. You’ll investigate how AI agents - spanning coding agents, SaaS AI platforms, cloud, and endpoints - can be abused, and design detections that stop threats before they cause impact. Each platform is different, yet they share common risk patterns across prompts, tools, MCP servers, skills, hooks, identity, and data movement. Your mission is to master both: uncovering unique attack surfaces while building universal, session- and intent-aware defenses. This role combines deep platform research, responsible security testing, and close collaboration with engineering to translate findings into production runtime protection. You will help define the field of Agentic Security.
Responsibilities
Research agentic threats: Prompt injection, tool/plugin/MCP abuse, skill and hook misuse, rogue autonomy, identity misuse, data leakage, adversarial AI, and protocol exploitation across coding agents and enterprise AI platforms.
Develop runtime detections: Design detections for multiple environments (coding agents, SaaS AI, cloud, endpoints), leveraging platform hooks, APIs, telemetry, and native guardrails where available.
Analyze platform complexities: Deep-dive into architectures, permissions, workflows, and agent capabilities (MCP servers, Skills, plugins, extensions) while extracting cross-platform risk patterns.
Build scalable detection logic: Prototype and refine heuristics, signatures, and intent-aware detectors that hold up at enterprise scale with low friction for builders.
Investigate monitoring and enforcement boundaries: Push what platforms allow for visibility and control - hooks, plugins, configuration surfaces, inventory signals, and policy enforcement points.
Prototype PoCs: Validate attack paths end to end, then turn findings into production-ready signatures, heuristics, detectors, and runtime policies.
Collaborate with engineering: Embed research into Dash’s detection pipelines, session-aware runtime protection, and policy/governance modules.
Stay ahead of the field: Track adversarial AI and agent security research, continuously evolving Dash’s detection and enforcement coverage.
Qualifications
BSc./MSc. in Computer Science, Security, or equivalent military/industry experience.
4+ years of security research (AppSec, malware, adversarial AI, or platform security). Experience with Agentic AI security research is an advantage.
Hands-on experience working with coding agents (e.g., Cursor, Claude Code) is a must, including practical use of agentic capabilities such as MCP servers, Skills, hooks, and similar agent tooling.
Experience with AI systems, agent frameworks, or protocol security (e.g., LangChain, MCP, A2A) is a strong plus.
Proficiency in Python for prototyping, research tooling, and detection logic; familiarity with TypeScript or Go is a plus for collaborating with engineering.
Strong grasp of cloud, SaaS, and endpoint security models, especially identity and data access.
Demonstrated ability to transform research into productized runtime detections.
Self-driven, curious, and eager to define the next frontier of enterprise security.