Frontier AI Offensive Security Specialist

Join Hitachi Cyber's Frontier AI Security practice and help organizations put their AI security controls to the test.

We are looking for an experienced penetration tester who wants to apply their expertise to one of the fastest-evolving areas of cybersecurity.

Our clients have integrated Large Language Models (LLMs) into customer-facing products, connected Retrieval-Augmented Generation (RAG) pipelines to sensitive data, and granted AI agents access to internal tools and business processes. While many have implemented security controls and guardrails to mitigate risk, they need experts who can determine whether those protections actually work.

As a Frontier AI Offensive Security Specialist , you will apply the same offensive security discipline used to assess web applications, APIs, cloud environments, and infrastructure to modern AI systems. Through threat modeling, adversarial testing, and adaptive attack campaigns, you will identify weaknesses, validate security controls, and provide clients with clear, evidence-based findings to help strengthen their defenses.

We are not looking for AI researchers. We are looking for experienced offensive security practitioners who want to apply their expertise to emerging AI technologies. Training on AI-specific technologies, frameworks, methodologies, and attack techniques will be provided.

Primary Responsibilities:

• Lead the scoping and rules of engagement for AI security validation assessments.
• Develop threat models for AI systems using frameworks such as MITRE ATLAS and the OWASP LLM Top 10, and collaborate with clients to define testing objectives.
• Conduct adversarial testing activities against AI-powered systems, including prompt injection, jailbreak, system prompt extraction, data exfiltration, tool abuse, RAG poisoning, and autonomous agent exploitation scenarios.
• Execute adaptive attack campaigns that evolve based on observed system behavior and validate the effectiveness of AI security controls and guardrails.
• Assess and document the effectiveness of AI security controls by determining whether they successfully block, partially mitigate, or can be bypassed by real-world attack techniques.
• Prepare and present technical findings, guardrail validation results, and prioritized remediation recommendations to client engineering and security leadership teams.
• Develop and maintain offensive testing methodologies, payload libraries, and retest procedures to support the efficient validation of remediation efforts.
• Collaborate with detection engineering teams to validate monitoring and detection capabilities through purple team exercises and real-world attack scenarios.
• Contribute to the continuous improvement of AI security tooling, testing frameworks, and AI-assisted security validation capabilities.

What You'll Work With:

• Advanced offensive security tooling, including Caido, Burp Suite, custom Python-based testing frameworks, and proprietary AI security assessment tools.
• AI-powered applications, LLM APIs, AI agents, vector stores, and AI guardrail technologies deployed across AWS, Azure, and GCP environments.
• Industry-leading frameworks and methodologies, including MITRE ATLAS, OWASP LLM Top 10, OWASP WSTG, OWASP MASTG, and NIST AI RMF.

Qualifications Required:

• 5+ years of hands-on experience conducting penetration testing engagements across web applications, APIs, mobile applications, cloud environments, or network infrastructures.
• Bachelor's degree in a technical field or equivalent demonstrated experience.
• Strong expertise in at least one offensive security domain: web/API (OWASP WSTG / ASVS), infrastructure and Active Directory, mobile (MASTG), or cloud configuration review.
• Demonstrated ability to identify and bypass security controls, such as web application firewalls (WAFs), input validation mechanisms, or content filtering solutions, and clearly articulate attack methodologies and findings.
• Experience performing source-assisted testing, including reviewing architecture documentation, code changes, or technical designs to identify attack paths and support security assessments.
• Practical scripting and automation experience, particularly with Python, and strong proficiency working in Linux-based environments.
• Demonstrated experience delivering client-facing reports and presenting findings to technical and business stakeholders.
• Excellent written communication skills and experience producing professional reports and technical deliverables.

Preferred qualifications (assets):

• Hands-on time attacking or defending LLM-backed applications (chatbots, copilots, RAG systems, autonomous agents).
• Familiarity with MITRE ATLAS, OWASP LLM Top 10 (2025), or NIST AI 600-1.
• Adversarial ML exposure (evasion, extraction, poisoning) from research, CTF, or production work.
• OSCP, GPEN, OSWE, CRT/CCT, or an equivalent practical offensive certification.
• Previous consulting or client-delivery experience is highly desirable.

We offer:

• Work at the forefront of AI security and innovation.
• Specialized training and mentorship on AI technologies, frameworks, and security methodologies.
• Collaboration with experienced offensive security, application security, detection engineering, and incident response professionals.
• A sense of belonging within a globally recognized cybersecurity organization.

Thank you for your application. Only selected candidates will be contacted.