Red Team Lead, AVP — AI Security & Adversarial Testing

Deutsche Bank Aktiengesellschaft; Madrid; Hybrid; EUR 120,000 - 180,000; Full time

Job summary

DWS Group is seeking a Red Team Lead, AVP in Madrid to lead adversarial testing of AI platforms and applications, identifying exploitable weaknesses and strengthening security controls. You will shape the AI security program and drive risk-based testing across the organization.

The role focuses on developing testing standards for AI/LLM environments, coordinating with engineering teams, and delivering actionable remediation plans. A strong background in offensive security within AI is essential.

Qualifications

• 8+ years of offensive security, penetration testing, or vulnerability research with hands-on AI/ML testing.
• Experience testing AI/ML, LLM, RAG, or agentic systems.
• Experience communicating vulnerabilities across applications, APIs, cloud platforms, data flows.
• Experience partnering with engineering teams through remediation and retesting.

Responsibilities

• Develop the firm’s AI red-team methodology and testing standards.
• Test for prompt injection, jailbreaks, data exfiltration, RAG manipulation, and agent/tool abuse.
• Conduct adversarial testing of AI use cases before production.
• Provide risk-rated findings and actionable remediation guidance.
• Track emerging AI attack techniques and incorporate into testing.

Skills

Offensive security

Penetration testing

Application security

Vulnerability research

AI/ML security testing

Tools

AI security tooling

Cloud security tooling

Job description

DWS Group is seeking a Red Team Lead, AVP in Madrid to lead adversarial testing of AI platforms and applications, identifying exploitable weaknesses and strengthening security controls. You will shape the AI security program and drive risk-based testing across the organization.

The role focuses on developing testing standards for AI/LLM environments, coordinating with engineering teams, and delivering actionable remediation plans. A strong background in offensive security within AI is essential.