Subject Matter Expert

About the role

You will define the strategic direction of child safety for Reflection, establishing the foundational principles that govern how the lab identifies and mitigates risks related to child sexual abuse and exploitation material. This role requires you to translate complex threat landscapes into concrete, actionable policies that can be implemented across open model lifecycles. You will own the entire evaluation lifecycle, from initial detection hypothesis to final enforcement decision, ensuring every release is rigorously assessed against the highest safety standards. As a critical gatekeeper, you will be responsible for maintaining the integrity of Reflection's open releases, ensuring they do not cause harm when deployed in real-world environments. You will directly manage the human and automated grading processes that determine model behavior, shaping the qualitative and quantitative metrics used to judge safety performance. Your work will involve deep collaboration with engineering and policy teams to close technical and procedural gaps before models are shipped to the public. You will also represent the organization in external forums, communicating our safety posture to regulators, researchers, and oversight bodies. Because the models are open, your judgment operates under unique constraints, requiring a high tolerance for ambiguity and a commitment to proactive risk identification. Ultimately, you will ensure that the lab's commitment to "intelligence for everyone" is balanced with robust protections for vulnerable populations.

Key facts

What you'll do

• Serve as the definitive internal subject matter expert on child safety, CSAM/CSEM, and online child sexual exploitation and abuse (CSEA) for Reflection AI.
• Establish and maintain the operational definitions, evaluation criteria, and escalation thresholds that govern the lab's approach to child safety enforcement.
• Design, implement, and scale automated evaluation pipelines and human review workflows to detect and mitigate risks in model outputs.
• Own the end-to-day management of the child safety review queue, including task prioritization, quality assurance, and SLA compliance.
• Act as the primary liaison and point of contact for all external content review partners, providing training, calibration, and performance feedback.
• Conduct detailed reviews of novel, ambiguous, or high-severity flagged content to inform enforcement actions and product decisions.
• Collaborate closely with the Safety, Alignment, Policy, and Engineering teams to integrate child safety considerations into pre-release evaluations and red-teaming exercises.
• Validate that all open model weights meet Reflection's child safety risk thresholds and operational standards prior to public deployment.
• Develop and maintain living documentation, including decision trees, playbooks, and standard operating procedures for scalable enforcement.
• Analyze and synthesize data on emerging misuse patterns, threat actor tactics, and adversarial behaviors to inform internal strategy.
• Partner with engineering to improve detection models, hash-matching algorithms, and automated enforcement mechanisms.
• Coordinate legal and compliance reporting obligations to bodies such as NCMEC in accordance with jurisdictional requirements.
• Monitor the evolving regulatory landscape, including KOSA, COPPA, and international frameworks, to ensure Reflection's policies remain current and effective.
• Represent Reflection's child safety expertise in external engagements with NGOs, academic researchers, and law enforcement agencies.
• Implement trauma-informed operational practices and safety protocols for team members engaged in reviewing sensitive content.
• Drive the continuous improvement of child safety metrics, ensuring they reflect the latest research and best practices in the field.

Requirements

• Possess deep subject matter expertise in child safety, child sexual exploitation and abuse (CSEA), or online child protection.
• Bring prior experience in trust and safety, content moderation operations, or policy enforcement involving child safety issues.
• Have direct experience managing content review operations, including queue management, quality assurance, and escalation procedures.
• Demonstrate a proven ability to scale policy enforcement and content review workflows in a high-stakes environment.
• Show proficiency in using SQL and data analysis tools to track workflow health, review metrics, and enforcement trends.
• Maintain a track record of identifying emerging risks and clearly communicating findings to cross-functional stakeholders, including Product, Policy, Engineering, and Legal.
• Exercise sound judgment when making high-stakes decisions under time pressure, particularly those impacting model release or termination.
• Demonstrate familiarity with established CSAM/CSEM classification standards, such as the COPINE Project scale or the SAM severity scale.
• Have direct experience working with or reporting to organizations like the National Center for Missing & Exploited Children (NCMEC) or the Internet Watch Foundation (IWF).
• Show a strong understanding of relevant legal and regulatory frameworks, including CSAM reporting laws, the Kids Online Safety Act (KOSA), and the Children's Online Privacy Protection Act (COPPA).
• Possess the ability to maintain resilience and professional composure when regularly exposed to explicit content involving minors, as well as violent or psychologically disturbing material.
• Commit to adhering strictly to Reflection's internal policies, external legal obligations, and the ethical guidelines governing open AI model releases.
• Be prepared to engage with distressing material as part of evaluating model outputs to ensure safety mechanisms are functioning correctly and comprehensively.
• Demonstrate a willingness to participate in ongoing training and calibration sessions to maintain consistency and accuracy in safety judgments.
• Understand the importance of discretion and confidentiality when handling sensitive case information and threat intelligence.

Nice to have

• Familiarity with the specific technical implementation of hash-matching and similar image recognition technologies.
• Experience contributing to the development of industry standards or best practices for child safety in open AI models.

Practical notes

This role may require occasional travel to meet with partners or attend industry conferences. Candidates must be eligible to work in the United States without sponsorship for this position. The position requires exposure to explicit content, and Reflection provides access to wellness resources and trauma-informed support to ensure team member wellbeing.