AI Red Teamer, Cybersecurity

This is a fully remote position, open to applicants in Washington.

📋 Description

• Assess the potential for AI models to be exploited for creating functional malware, effective exploit code, attack tools, or operational cyberattack strategies.

• Create adversarial prompts and multi-turn interaction scenarios that mimic threat actors through the phases of reconnaissance, weaponization, exploitation, lateral movement, persistence, and exfiltration.

• Determine if model-generated payloads, exploit sequences, and attack strategies are genuinely hazardous and practically applicable.

• Develop technically sound adversarial prompts that align with the cyber kill chain.

• Review model-generated code and technical outputs for their functional accuracy.

• Conduct tests on malware generation, vulnerability exploitation, social engineering, credential harvesting, privilege escalation, command-and-control infrastructure, and data exfiltration.

• Explore the dual-use boundaries concerning security research, penetration testing, and defensive operations.

• Simulate a range of attackers from opportunistic to advanced persistent threat actors.

• Evaluate multi-step and multi-turn attack sequences.

• Assess model responses using structured harm taxonomies and severity evaluation rubrics.

• Document findings with comprehensive technical explanations.

• Develop and enhance cybersecurity evaluation frameworks and threat models.

• Collaborate with red team members, AI researchers, and policy teams to convert findings into model enhancements.

• Keep abreast of offensive-security tactics, techniques, and procedures (TTPs), CVEs, jailbreaks, and advancements in AI/offensive security.

⛳️ Requirements

• Professional experience in offensive security, penetration testing, red teaming, vulnerability research, malware analysis, threat intelligence, or incident response.

• Proficiency in reading, writing, and evaluating code in languages such as Python, PowerShell, Bash, C/C++, or JavaScript.

• A solid understanding of MITRE ATT&CK and OWASP frameworks.

• Capability to evaluate the functional accuracy and real-world exploitability of model-generated technical outputs.

• Extensive hands-on experience with various large language models (LLMs), including ChatGPT, Claude, Gemini, or open-source alternatives.

• Innovative and adversarial problem-solving abilities.

• Clear and effective written communication skills, including the ability to convey technical risks to non-specialist audiences.

• Strong ethical judgment.

• Capability to work independently while effectively collaborating in a feedback-oriented, distributed environment.

• Relevant certifications such as OSCP, OSCE, GPEN, GXPN, CRTO, CRTL, CEH, or similar are preferred.

• Active or previous security clearance is a plus.

• Experience in exploit development, reverse engineering, or binary analysis is advantageous.

• Background in cloud security, container security, or infrastructure-as-code attack surfaces is desirable.

• Familiarity with AI/ML attack surfaces, including prompt injection, model extraction, training-data poisoning, and adversarial examples is a bonus.

• Experience in constructing or operating command-and-control frameworks, custom implants, or offensive tools is a plus.

• A record of bug bounties or published CVEs is a plus.

• Previous experience in trust and safety, content moderation, or AI evaluation is beneficial.

• Familiarity with LLM APIs or evaluation tools is an advantage.

• Candidates must have the right to work legally in the United States; the application will inquire if visa sponsorship is needed.

• Ability to engage with offensive cybersecurity content in a professional and responsible manner.

🏝️ Benefits

• Flexible work arrangements, including part-time options.

• Opportunity to work remotely from any location within the United States or from Seattle.

• Collaborative efforts through virtual working sessions.