AI Safety Jobs

Research Engineer Irregular Tel Aviv, Israel In this role, you'll build systems to evaluate and secure frontier AI models at Irregular. Develop infrastructure and experiments to assess model capabilities and implement agent frameworks. Create robust evaluation pipelines and security-focused testing frameworks. Design challenges to measure models' ability to evade detection by defensive security tools. Build controlled environment frameworks and tools that help understand and mitigate risks related to frontier models. Irregular Labs is a frontier security lab that aims to protect the world in a time of increasingly capable and sophisticated AI...

Technical Policy Researcher Irregular Tel Aviv, Israel In this role, you'll tackle complex questions of AI development and deployment as a Technical Policy Researcher, shaping the emerging field of AI security. Conduct threat modeling to identify specific ways strong cyber-capable models could cause harm. Develop taxonomies for dangerous AI capabilities and potential mitigations. Create policy proposals for models' refusal policies and track developments in AI policy and research. Write academic papers and blog posts detailing research on evaluation theory and mitigation recommendations. Irregular Labs is a frontier security lab that aims to ...

AI Red Team Analyst Alignerr Remote, Global $15 - $75 per hour In this role, you'll conduct red-teaming exercises to uncover AI security weaknesses and deliver findings that improve system safety. Craft adversarial prompts, jailbreak attempts, and edge-case scenarios to challenge AI model guardrails. Evaluate AI outputs for safety violations, bias, and policy compliance. Document vulnerabilities and unexpected behaviours in structured reports for engineering teams. Collaborate with teams to recommend security mitigations and help refine testing protocols. Alignerr is a platform that for people to create data for frontier AI labs.

Join Reflection AI as a Member of Technical Staff - Data Quality Engineer to own upstream data quality and automated QA methods (including LLM-as-a-Judge frameworks) for LLM post-training and evaluation.

Machine Learning Researcher Gray Swan Remote, USA, Remote, Global In this role, you'll conduct AI safety research to identify and defend against failure modes in advanced AI systems. Discover emerging failure modes through stress-testing cutting-edge AI models. Help enterprises deploy AI safely and at scale without compromising innovation. Inform official safety evaluations of the world's most advanced AI models. Develop research-backed solutions for emerging AI safety challenges. Gray Swan is an AI security company that develops tools that automatically assess the risks of AI models.

Senior communications role focused on external messaging for OpenAI safety research, including alignment, evaluations, preparedness, and interpretability.

The role involves pursuing a PhD with a focus on AI safety, generative AI, and agentic AI systems, addressing critical aspects of AI development and deployment.

Research engineer building AI platforms and infrastructure for alignment research, safety research, and evaluation of AI alignment/control approaches at CARMA.

The AI Red Teamer role involves probing frontier AI systems for vulnerabilities, designing attack strategies, and building safety evaluation infrastructure to ensure AI safety.

Research Scientist role studying latent structure and behavior in neural networks, with explicit emphasis on understanding as safety, safety-relevant tools, and internal red teaming.

Technical lead for ARIA’s multi-agent security programme, supporting research on AI agents operating and coordinating in untrusted environments, with explicit emphasis on AI red-teaming and security research.

Computational biologist role focused on building evaluation frameworks and experiments for biological security problems using frontier AI systems.

Role building evaluation task pipelines and infrastructure for frontier model testing, including systems to prevent models from detecting evaluations.

Product engineer building interfaces, APIs, and workflows that operationalize interpretability research for training, evaluating, debugging, and deploying AI systems; relevant because it supports safety-related understanding of model internals and safer AI systems.

Funding call for research on AI safety in multi-agent systems, including testbeds, agent networks, infrastructure, and oversight/control for frontier-model agents.

Lead a research program on Provable AI Safety, focusing on building AI systems that verify their own correctness and collaborating with teams to advance AI safety.

The AI Security Systems Architect will design and develop systems for security testing and evaluation of AI technologies, focusing on adversarial testing methodologies and ensuring the resilience and security of AI systems.

Research scientist role focused on designing and running evaluations of AI harmful manipulation for EU AI Act enforcement, including frontier model evals, red-teaming, and regulator-facing reporting.

Technical project manager for a research division building harmful-manipulation evaluations for the EU AI Office, coordinating delivery, budgets, partners, and hiring for AI safety evaluation work.