AI Red Team Engineer

Arcitix AI Security; United States; Remote; USD 120,000 - 180,000; Full time

Job summary

Arcitix AI Security seeks a security-focused researcher to assess abuse vectors in language models and AI agents. You will explore jailbreaks, prompt injections, data exfiltration risks, and attack chains to help build safer systems.

The role requires hands-on testing of ML systems, documenting weaknesses and suggesting mitigations. You will collaborate with engineering teams to translate findings into practical security controls.

Qualifications

• Ability to evaluate the security of language models, agents, or other machine learning systems.
• Familiarity with jailbreaks, prompt injection, data exfiltration, and AI abuse scenarios.

Responsibilities

• Probe systems for jailbreaks and ways to bypass intended safeguards.
• Test how models and agents respond to prompt injection.
• Investigate potential paths for sensitive data exposure or exfiltration.
• Identify abuse patterns that combine multiple weaknesses into a broader attack chain.

Skills

Security assessment

AI/system vulnerability testing

Job description

Role overview

Assess how language models, AI agents, and machine learning systems can be manipulated or abused. The work focuses on uncovering security weaknesses and attack paths before they can be exploited.

Responsibilities

• Probe systems for jailbreaks and ways to bypass intended safeguards.


• Test how models and agents respond to prompt injection.


• Investigate potential paths for sensitive data exposure or exfiltration.


• Identify abuse patterns that combine multiple weaknesses into a broader attack chain.

Requirements

• Ability to evaluate the security of language models, agents, or other machine learning systems.


• Familiarity with jailbreaks, prompt injection, data exfiltration, and AI abuse scenarios.