Jobs at Alice (Formerly ActiveFence)

Lead AI safety research across generative AI models, multimodal systems, and AI agents. Design scalable evaluation and red-teaming methodologies, investigate jailbreaks and prompt injection, assess content safety risks, manage multilingual datasets, document research findings, mentor junior researchers, and communicate mitigation recommendations to engineering, product, policy, and client teams.

Analyze generative AI content infringements and develop adversarial prompts to identify model vulnerabilities across hate speech, misinformation, intellectual property, and other abuse areas. Manage multilingual datasets, investigate safety circumvention tactics, oversee projects and quality assurance, and collaborate with engineering, product, and policy teams to improve AI safety strategies.

Lead a multidisciplinary red teaming team to design and run adversarial tests, evaluate model risks, produce high-quality deliverables, and communicate findings to clients while improving methodologies and workflows.

Lead AI safety role developing adversarial prompt strategies, owning end-to-end projects, mentoring junior analysts, managing multilingual abuse datasets, researching circumvention tactics, and partnering with engineering, product, and policy teams to secure generative AI models.

Analyze content infringements and write adversarial prompts to find model vulnerabilities across LLMs, text-to-image/video, and agents. Manage datasets and projects end-to-end, investigate evasion tactics, collaborate with engineering, product, and policy teams, and promote knowledge sharing to improve model safety.

Adversarially test and strengthen Generative AI models against cyber-CBRNE threats by designing red-team prompts, jailbreaks, multi-turn attacks, and scenario-based evaluations; develop test suites, taxonomies, and actionable mitigations while documenting vulnerabilities and maintaining threat and AI-safety expertise.

Evaluate and strengthen safety guardrails for generative AI by adversarial red-teaming, prompt research, and risk assessments. Translate biological expertise and regulatory standards into actionable model-safety feedback, monitor emerging biotech threats, and update biosecurity benchmarks.

Evaluate and strengthen safety guardrails for generative AI handling chemical information through adversarial red-teaming, taxonomy audits, model pairing, and risk identification. Identify dual-use risks, clandestine synthesis instructions, and toxicological inaccuracies. Maintain regulatory and international chemical convention knowledge and communicate findings with clear technical reports to inform safety and alignment decisions.