Jailbreaking Lead (Red Team)

Far AI, Inc.

Job Overview

Location

Remote (International)

Job Type

Full-time

Full Job Description

📋 Description

• Lead hands-on jailbreaking of frontier AI models, including closed-weight and open-weight systems, focusing on high-severity, universal, and near-universal vulnerabilities across CBRN, cyber, agentic security, extreme persuasion, and emerging risk domains.
• Systematically dismantle defense-in-depth stacks by chaining novel and established attack techniques through input filters, model-level refusal mechanisms, reasoning monitors, output filters, and account-level moderation systems.
• Escalate initial vulnerabilities to their most severe, universal forms, maximizing success rate, reliability, and the capability of elicited harmful outputs.
• Invent new attack classes when existing methods fail, including against advanced defenses like Constitutional Classifiers and fine-tuning APIs, and rapidly incorporate state-of-the-art techniques from research literature.
• Set the technical standard for vulnerability severity and generality across all major red-teaming engagements, defining what constitutes a credible, impactful jailbreak.
• Spend 50–70% of time personally conducting jailbreaks, with remaining time dedicated to mentoring ICs, conducting pairing sessions, retrospectives, and internal writeups to scale team capability.
• Review and validate red-teaming deliverables for technical accuracy, severity judgment, and clarity, ensuring findings meet the highest bar for impact and precision.
• Work directly with leading AI labs (e.g., OpenAI, Anthropic) and government agencies (e.g., EU AI Office, UK AI Security Institute) to translate vulnerabilities into real-world mitigations, not just disclosures.
• Contribute to public benchmarks, safety leaderboards, and reports that shape industry norms and government policy around AI safety.
• Make calibrated technical judgments on what vulnerabilities are universal, reliable, and exploitable by capable threat actors—prioritizing those with real-world consequences.
• Shape the jailbreaking research agenda in partnership with leadership, ensuring FAR.AI’s toolkit evolves ahead of model defenses and emerging model affordances (e.g., agents, multimodal, long context).
• Stress-test novel AI system capabilities as frontier models advance, including tool use, reasoning chains, and multimodal inputs, to identify new failure surfaces.
• If on the management track: hire, manage, and grow a jailbreaking team while maintaining top-tier personal technical performance and hands-on output.
• Maintain deep understanding of LLM architectures, training processes, and failure modes, and how these influence adversarial behavior under defensive constraints.
• Operate in fast-moving, ambiguous environments where attack surfaces change weekly, requiring constant innovation, adaptation, and relentless pursuit of new jailbreaks.
• Communicate technical findings effectively to both technical teams at AI labs and non-technical policymakers, ensuring findings drive action and systemic improvements.
• Publish and speak honestly about risks without constraint, contributing to the public AI safety discourse through transparent reporting and community engagement.
• Thrive in a mission-driven culture where success is measured by adopted mitigations, shifted industry standards, and reduced real-world harm—not academic publications or internal metrics.
• Operate with independence, autonomy, and a “get shit done” attitude, doing whatever is necessary to achieve impact in a lean, high-velocity organization.
• Navigate uncertainty and shifting threat models with speed, creativity, and precision, consistently outpacing defensive improvements by frontier AI developers.
• Collaborate with a global red team empowered by proprietary tools, research infrastructure, and strategic partnerships to maximize the scale and impact of your work.

🎯 Requirements

• Personally developed universal or near-universal jailbreaks against at least one leading frontier AI model
• Demonstrated ability to chain multiple attack techniques through defense-in-depth stacks in heavily protected AI systems
• Deep hands-on experience with black-box optimization, multimodal attacks, and/or agentic red-teaming
• Strong track record in AI, adversarial machine learning, cybersecurity, or a highly technical field (e.g., computer science, math, physics)
• Proven ability to invent novel attack classes and adapt quickly as model defenses evolve
• Clear demonstration of relentless drive to achieve high-impact results in ambiguous, fast-moving environments

🏖️ Benefits

• Compensation range of USD 170,000–250,000, with potential for more for exceptional candidates
• Remote work globally with visa sponsorship available for the USA or Singapore
• Full-time position with up to one international trip per month for convenings, government meetings, or team gatherings
• Opportunity to work with leading AI labs (OpenAI, Anthropic), government agencies (EU AI Office, UK AI Security Institute), and top researchers (e.g., Yoshua Bengio)
• Mission-driven culture focused on real-world impact, with autonomy to publish and speak openly about AI risks
• Access to FAR.Labs, a Berkeley-based AI safety co-working space, and the broader AI safety research community through grants and events

Skills & Technologies

REST

Senior

Remote

Ready to Apply?

Apply Externally

You will be redirected to an external site to apply.

AI Job Fit Analysis

Pro

See exactly how your profile matches this role — strengths, skill gaps, and what to do about them.

Far AI, Inc.

Visit Website

About Far AI, Inc.

Far AI is a technology company focused on developing advanced artificial intelligence solutions for the logistics and supply chain industry. Their core product is an AI-powered platform designed to optimize complex logistical operations, including route planning, warehouse management, and demand forecasting. By leveraging machine learning and data analytics, Far AI aims to enhance efficiency, reduce costs, and improve the overall resilience of supply chains for their clients. The company serves a diverse range of industries that rely heavily on efficient transportation and inventory management, positioning itself as a key player in the digital transformation of logistics.

View Company Profile

Get more remote jobs like this

Subscribe to the weekly newsletter for similar remote roles and curated hiring updates.

Weekly remote jobs and featured talent.

No spam. Only curated remote roles and product updates. You can unsubscribe anytime.