Security Controls Assurance Lead

About Anthropic Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems. About the role Anthropic's Security Governance, Risk, and Compliance (GRC) team is the connective tissue that holds the company accountable to its security commitments. We translate regulatory, customer, and voluntary obligations into controls that teams act on, and give leadership a bird's-eye view of how well we're meeting them. We're building toward a fundamentally different kind of GRC: one that directs Claude, with the right humans in the loop, to challenge and evidence the performance of controls continuously rather than through periodic audits. We are designing an integrated compliance and risk ecosystem that serves as a trust engine and an independent risk advisor as Anthropic governs itself at a level beyond frameworks. As part of Security GRC's technical controls assurance function, you will be the voice on what the control environment must achieve. You will define control requirements and acceptance criteria for our global compliance obligations (e.g. SOC 2, ISO 27001/42001, HIPAA, public sector) across the software development lifecycle, pair with engineering as they design and implement against those requirements, and validate that what ships actually meets the bar. Key responsibilities - Define the control framework and requirements for autonomous AI operators in collaboration with Security, Internal Audit, and Engineering, including change review and approvals, human-in-the-loop, and evidence collection. Assess implementations against those requirements. - Pressure-test major infrastructure, system, and agent framework changes for control impact during design, before decisions become expensive rework. - Set the compliance bar for home-built systems. Collaborate with teams to define what the internal system must provide from day one, such as auditability, segregation of duties, and change control over the tool itself. - Define the criteria for where and when AI can operate, supplement, or replace a manual process or control, including the human-in-the-loop thresholds and evidence documentation. - Establish the validation, evidence, and governance standards that allow AI-performed and AI-assisted processes and controls to withstand external audit and regulatory scrutiny. - Assess the introduction of new compliance frameworks and changes in scope (new regulations, certifications, products, or entities), providing a sufficient technical and compliance lens on their impact to control design, evidence requirements, and engineering effort before commitments are made. - Stand up or advise on audit workflows for the assurance team, including Claude-driven control testing, automated evidence collection, walkthrough preparation, and framework mapping against our common controls framework, materially raising automated evidence coverage and cutting audit prep time. Minimum qualifications - Thrive at the pace of a hypergrowth company. You’re comfortable making calls with incomplete information and reprioritizing as scope shifts. - Have supported technology control programs through SOX readiness or as a public company or with equivalent rigor (FedRAMP, large multi-framework SOC 2/ISO portfolios). - Have genuine engineering fluency, possibly from an earlier engineering career: you can read code and Terraform, follow a CI/CD pipeline end to end, and challenge a design on its technical merits. - Have programming skills in Python or at least one systems language such as Go, Rust, or C/C++. - Have deep familiarity with developer platform, release engineering, or infrastructure control domains. - Are a strong collaborator and communicator. - Use Claude and other LLMs as daily working tools, and have grounded, specific views on which audit and assurance workflows AI can run today and which it can't yet. - Translate framework and regulatory language into acceptance criteria engineers can build against, and translate engineering reality back into assurance language auditors and leadership can rely on. - Default to getting the requirement designed into the system rather than papering over the gap with procedure. Preferred qualifications - Have a combination of audit or advisory experience (Big 4 or equivalent) with in-house experience at an AI-forward tech company — in either order - Have defined or assessed controls for AI/ML systems or agents acting in production environments - Have stood up continuous controls monitoring or automated evidence programs Candidates need not have - Done everything on this list. This role does not require writing production code day to day. We encourage and expect you to ship, but the bar is fluency sufficient to

Thrive at the pace of a hypergrowth company. Comfortable making calls with incomplete information and reprioritizing as scope shifts. Supported technology control programs through SOX readiness or as a public company or with equivalent rigor (FedRAMP, large multi-framework SOC 2/ISO portfolios). Genuine engineering fluency: can read code and Terraform, follow a CI/CD pipeline end to end, and challenge a design on its technical merits. Programming skills in Python or at least one systems language such as Go, Rust, or C/C++. Deep familiarity with developer platform, release engineering, or infrastructure control domains. Strong collaboration and communication. Ability to use Claude and other LLMs as daily working tools, with grounded views on which audit and assurance workflows AI can run today and which it can\'t yet. Ability to translate framework and regulatory language into acceptance criteria engineers can build against, and translate engineering reality back into assurance language auditors and leadership can rely on. Default to getting the requirement designed into the system rather than papering over the gap with procedure.

Define the control framework and requirements for autonomous AI operators in collaboration with Security, Internal Audit, and Engineering, including change review and approvals, human-in-the-loop, and evidence collection. Assess implementations against those requirements. Pressure-test major infrastructure, system, and agent framework changes for control impact during design, before decisions become expensive rework. Set the compliance bar for home-built systems. Collaborate with teams to define what the internal system must provide from day one, such as auditability, segregation of duties, and change control over the tool itself. Define the criteria for where and when AI can operate, supplement, or replace a manual process or control, including the human-in-the-loop thresholds and evidence documentation. Establish the validation, evidence, and governance standards that allow AI-performed and AI-assisted processes and controls to withstand external audit and regulatory scrutiny. Assess the introduction of new compliance frameworks and changes in scope (new regulations, certifications, products, or entities), providing a sufficient technical and compliance lens on their impact to control design, evidence requirements, and engineering effort before commitments are made. Stand up or advise on audit workflows for the assurance team, including Claude-driven control testing, automated evidence collection, walkthrough preparation, and framework mapping against our common controls framework, materially raising automated evidence coverage and cutting audit prep time.