Operations Specialist (Red Team Evaluation) – TikTok Safety Products

Written by the T&S Network from TikTok's posting.

Red team evaluation of global content security systems, testing them from an attacker's perspective to find vulnerabilities and recommend fixes across cultural and linguistic contexts. Covers risk assessment frameworks, threat hunting, and AI-powered risk assessment workflows for TikTok's safety products. Sits within the Safety Products team.

• AI misuse and jailbreaks

What you'd do

• Shape and refine risk assessment processes, workflows, and SOPs spanning signal detection, escalations, and investigations.
• Maintain the Red Team Risk Taxonomy and forecast emerging risks from regulation, new technology, and AIGC trends.
• Run proactive threat hunts and in-depth risk investigations using enforcement gaps, escalations, sentiment, and abuse intelligence.
• Define agent capabilities, prompt designs, and Human-in-the-Loop frameworks for the Agentic Risk Assessment Platform with engineering and ML teams.
• Stand up and run threat intelligence relationships with security vendors, white-hat groups, and industry peers.
• Translate risk findings into mitigations with Product, Policy, Data Science, T&S, and Operations teams.

What it asks for

• Demonstrated work in T&S risk assessment, investigations, policy enforcement, content moderation strategy, or online abuse mitigation.
• Alternatively, background in CTI, red teaming, blue teaming, penetration testing, or security operations.
• Working understanding of LLM and multimodal AI capabilities and limits, with hands-on prompt design and human-in-the-loop workflow experience.
• Proficiency in SQL and Python, with experience building investigation dashboards and measuring metrics like enforcement miss rates.
• Required proficiency in Chinese to communicate with internal stakeholders in China.
• Experience assessing multimodal and AIGC risks such as deepfakes and synthetic audio or video (preferred).

Terms in this role

Red teaming.

Testing a system, often an AI model, by deliberately trying to make it fail or misbehave, so problems are found and fixed first.

Our best summary of the posting, and it can be wrong. The employer's posting is what counts: read it in full on TikToknorth_east before you apply. If we got something wrong, report this listing.