Speculative Research Engineer, Red Team Expeditionary Lead
About Anthropic
Anthropic’s mission is to build reliable, interpretable, and steerable AI systems that are genuinely beneficial to society. We are a rapidly growing group of researchers, engineers, policy practitioners, and operators working together to make frontier AI safer in practice, not just in principle.
About the Team
The Frontier Red Team is a small technical research team focused on understanding and defending against the risks posed by increasingly autonomous AI systems. We build evidence, not just arguments: model organisms, concrete demonstrations, and field-ready defenses that clarify what matters before stakes escalate.
This role leads an expeditionary research mode inside red teaming. You will design speculative but reality-anchored prototypes that expose vulnerabilities, incentives, and system dynamics that conventional testing misses. You will convert those artifacts into rigorous evaluations, adversarial drills, and engineering interventions that can be adopted across teams.
We are especially interested in candidates who can move fluently across technical systems and human systems: people who can code deeply, reason historically, communicate clearly, and build things that change decisions.
What You'll Do
- Design and implement speculative model organisms of autonomous systems to surface high-consequence failure pathways before they are obvious in telemetry.
- Build adversarial evaluations and training environments that pressure-test agent behavior across technical, organizational, and social contexts.
- Develop defensive agents and workflows that can detect, disrupt, or outcompete adversarial autonomous systems in realistic scenarios.
- Create artifact-led demonstrations (prototypes, scenarios, traces, walkthroughs) that make frontier risks legible to leadership, policy, and external stakeholders.
- Pair with research, engineering, and policy teams to turn findings into concrete mitigations, launch criteria, and governance inputs.
- Lead fast iteration cycles from ambiguous question to reproducible evidence to deployable safeguard.
Sample Projects
- Build a multi-agent simulation where a cooperative assistant drifts into strategic misalignment under subtle interface and incentive changes, then ship an evaluation harness that catches the drift early.
- Prototype a cyber-physical tabletop where language agents coordinate software and hardware actions; derive concrete defense patterns and escalation triggers.
- Construct a “possible-futures lab” pipeline that turns weak sociotechnical signals into red-teamable artifacts and prioritized experiment backlogs.
- Develop a rapid-response demo for a policy audience that shows how a plausible autonomy failure unfolds end-to-end and how mitigation changes the outcome.
You May Be a Good Fit If You
- Have strong engineering capability in Python and modern AI tooling, and can ship robust research code quickly.
- Have hands-on experience with LLM agents, autonomous systems, adversarial testing, or security evaluation.
- Can use imagination as an engineering instrument: you generate unconventional but testable hypotheses and pursue them rigorously.
- Have a track record of building things outside narrow big-tech lanes: startups, labs, independent research, open communities, or interdisciplinary programs.
- Can produce compelling technical artifacts that move cross-functional decisions, not just internal technical debate.
- Bring strong written communication and narrative clarity; you can explain consequential technical risk to mixed audiences.
- Value collaborative practice, including pair-programming, shared authorship, and community-oriented leadership.
Strong Candidates May Also Have
- Experience in reinforcement learning, multi-agent systems, self-play, or evaluation infrastructure for frontier models.
- Backgrounds that broaden technical sightlines: history of technology, STS, design research, anthropology, media practice, or similar.
- Serious creative practice (e.g., photography, writing, filmmaking, prototyping, archives, maker culture) that strengthens your research method.
- Entrepreneurial experience with end-to-end ownership under uncertainty.
- Experience collaborating with external partners across policy, public interest security, academia, or civil society.
Annual Salary:
The annual compensation range for this role is listed below.
$350,000 - $850,000 USD
Logistics
- Education requirements: Bachelor’s degree in a related field or equivalent practical experience.
- Location-based hybrid policy: We currently expect staff to be in an Anthropic office at least 25% of the time; some projects may require more in-person collaboration.
- Visa sponsorship: We do sponsor visas when feasible for the role and candidate, and we make reasonable efforts to support immigration pathways when extending an offer.
- Relocation: We are open to relocation for this role and assess case-by-case support.
How we're different
We work as one team on a small number of high-consequence research efforts. We prioritize impact over novelty theater: if a prototype does not improve decisions, mitigations, or readiness, it is not done.
The easiest way to evaluate fit is to look at your own portfolio and ask: where have you already used making as a way of thinking under uncertainty? We care about that deeply.
Come work with us!
Anthropic is a public benefit corporation headquartered in San Francisco. We offer competitive compensation and benefits, optional equity donation matching, generous vacation and parental leave, flexible working hours, and a collaborative office environment.
Guidance on Candidates' AI Usage: Learn about our policy for using AI in our application process.
We encourage you to apply even if you do not meet every listed qualification. Strong candidates come from many routes, and we value a broad range of lived, professional, and disciplinary perspectives.
Your safety matters to us. Anthropic recruiters only contact candidates from @anthropic.com addresses or clearly identified partner agencies. We will never request fees or banking information prior to employment.
As set forth in Anthropic’s Equal Employment Opportunity policy, we do not discriminate on the basis of any protected group status under applicable law.
Create a Job Alert
Interested in building your career at Anthropic? Get future opportunities sent straight to your email.
Create alertApply for this job
*
indicates a required field