Staff + Software Security Engineer, Labs

Anthropic is building AI systems designed to be reliable, interpretable, and safe for society. As frontier AI capabilities advance rapidly, the security challenges grow more severe and adversarially complex than anything the industry has faced before. This Staff Security Engineer role joins Security Labs, a specialized team tasked with exploring high-risk, high-impact research bets that tackle the most difficult security problems blocking safe AI deployment at scale.

Security Labs operates differently from traditional security teams. You'll work on experimental projects with short timelines—measured in weeks rather than quarters—where roughly one-third of bets are expected not to land. Each project either gets handed to an Anthropic team for production ownership or closes with documented learnings. This environment demands engineers who thrive in ambiguity, move quickly through uncharted territory, and find experimental failure as valuable as success.

What you'll do

• Own security research projects from conception through execution, including scoping the bet, building working prototypes, testing against real training and inference workloads, and deciding between handoff or documented shutdown
• Design and stand up novel security infrastructure rapidly—isolated compute clusters, cryptographic attestation chains, hypervisor-based workload isolation, and verification tooling—optimized for learning rather than permanence
• Run experiments testing whether research and infrastructure workflows can survive proposed security controls, documenting exactly where friction points emerge
• Partner with Pretraining, RL, Inference, and Compute teams to validate that your prototypes integrate with their actual work
• Translate experimental results into architectural recommendations and actionable contingency plans that shape how Anthropic builds secure systems long-term
• Help select future research bets and contribute to raising security standards across the industry

What you'll bring

• Genuine conviction that AI trajectory matters deeply and commitment to solving the security problems that determine whether it develops safely
• Proven depth in at least one specialty most software engineers never develop: firmware or hardware security, applied cryptography, OS/kernel/hypervisor internals, formal methods and verification, reverse engineering, or high-assurance cross-domain systems
• Track record of building and shipping things independently—founding a company or research group, maintaining critical open-source projects, or publishing research that influenced your field
• Comfortable jumping between abstraction layers when the problem demands it and willing to attempt ambitious approaches that might fail rather than iterate safely
• Ability to move fast, communicate findings clearly in writing, and work productively without extensive structure or certainty

Nice to have

• Background in distributed systems or large-scale infrastructure security
• Experience with formal verification or machine-checked proofs
• Prior work on cryptographic protocols or zero-knowledge systems

What they offer

• Full-time position based in San Francisco, New York City, or Seattle
• Work on some of the most consequential security challenges in AI development