Daniel Olmedilla

Distinguished Engineer, Trust and Safety, LinkedIn

Title: Scaling Abuse Prevention for the Next Wave of AI at Linkedin

Abstract: AI is reshaping the abuse landscape by enabling attackers to generate more convincing content, automate workflows, and iterate at unprecedented speed. At the same time, platforms are increasingly using AI themselves, creating both new opportunities for defense and new classes of failure modes. As organizations build more sophisticated AI solutions and agentic experiences, abuse prevention must evolve in parallel to address risks that are more dynamic, scalable, and difficult to contain. This talk explores how LinkedIn is scaling abuse prevention for this changing environment, including the challenges of detection and enforcement in AI-mediated ecosystems, the need to embed trust and safety into advanced AI systems from the start, and how to build sophisticated AI-powered defenses and agents that can detect, adapt, and respond at scale, , while continuously improving decision quality, agility, and operational efficiency.

Bio: With over 20 years of work experience in diverse domains and industries, Daniel is a seasoned leader and innovator in the field of artificial intelligence, data science and product infrastructure, having managed large teams across multiple sites world-wide. His mission is to support machine learning efforts at LinkedIn to ensure a safe, trusted, and professional platform, while committing to the advancement of AI driven by ethical principles that put people first. He brings diverse perspectives and experiences to the team, as a multilingual professional with a PhD in computer science and a background in industry, consulting, academia, and research.

Baraa Hamodi
Principal Software Engineer, Trust and Safety, Meta

Title: Agentic AI Integrity Support - Conversational Assistance for Content Moderation and User Remediation at Scale

Abstract: Users impacted by integrity enforcement actions on social media platforms face significant barriers to understanding policies, navigating remediation options, and exercising their right to appeal. Traditional support surfaces are fragmented across multiple entry points and offer limited, static assistance — resulting in low satisfaction and under-utilization of remediation pathways. We present an agentic AI support system deployed on Facebook and Messenger that provides multi-turn, context-aware conversational assistance for users affected by content moderation enforcement. The agent leverages personalized enforcement history to deliver transparency into specific policy violations, explain community standards in natural language, and surface tailored remediation options including in-chat appeals. Deployed across Account Status and Meta AI, the system was developed through cross-organizational collaboration across Facebook and Instagram. In production evaluation, the agent demonstrated significant improvements in user sentiment across both general support and integrity-specific intents, while meaningfully increasing engagement with appeal workflows compared to traditional support paths. We discuss architectural decisions for grounding agent responses in enforcement context, the phased approach to scaling from individual use cases to platform-wide support, and implications for human-AI collaboration in integrity operations.

Bio: Baraa is a Principal Software Engineer at Meta, where he currently drives the vision and strategy for AI-powered developer productivity - architecting Agent-AKI tools, workflows, and platforms that are transforming how engineers build software at scale. Prior to this role, Baraa spent eight years leading Trust and Safety Initiatives at Meta, where he designed and built Integrity Systems and user-facing safety products that protect billions of people across Meta's family of apps. His works span the detection and mitigation of platform abuse, misinformation, and harmful content, contributing to the development of scalable enforcement and content moderation infrastructure. Baraa also founded cross-functional engineering communities at Meta focused on raising the bar for product quality and engineering excellence. He is passionate about the intersection of artificial intelligence, platform integrity, and building systems that operate reliably and responsibly at scale. 

Karthik Lakshmanan
Principal Engineer/Director, YouTube Trust and Safety, Google

Title: Overview of AI-models for Trust and Safety at YouTube

Abstract: In this talk, I’ll provide an overview of how we use AI to detect policy-violative content on YouTube across different entity types: videos, comments, livestreams, etc and keep our community safe. AI is an essential part of YouTube Trust and Safety and I'll share the lessons we learned over many years while building, deploying and maintaining models at scale. Furthermore, I’ll cover the unique challenges for different product surfaces, like livestreams and shorts, which makes the problem domain particularly intriguing.

Bio: Karthik Lakshmanan received his Ph.D. in Electrical and Computer Engineering from Carnegie Mellon University, in 2011. He is now a Principal Engineer / Director at Google, working primarily on Trust and Safety, GenAI Safety, and Large Language Modeling at YouTube. His current research interests include Natural Language Processing, Content Understanding, and Building Scalable Classification Systems. Previously, he also worked on Large-scale Recommenders, Embedding Models, and Real-Time Scheduling theory for Embedded (of a different kind) Systems. 

Anish Das Sarma

Founder & CEO, Reinforce Labs

Title: AI Integrity as a Search Problem: Diversity-Driven Behavioral Evaluation

Abstract: It has become too easy to generate and too hard to evaluate. AI systems ship fast, but making sure they actually behave correctly is still slow, manual, and narrow. Current evaluation makes this worse - it optimizes for one number (attack success rate) and misses the breadth of how systems really break. We present Flint, a framework that treats the gap between how AI behaves and how it should behave as a search problem. Flint is a search engine - the target system is just a parameter. The same infrastructure that probes a chatbot's safety boundaries works for agent frameworks, RL reward models, and world model evaluation. Only the executor adapter changes. Flint nests evolutionary strategy search with RL-guided prompt optimization. The outer loop evolves multi-turn, multi-modal strategies and maintains a quality-diversity archive that keeps diverse high-performers rather than collapsing to one best approach. The inner loop sequences mutations turn by turn, guided by a belief tracker (BeliefNet) that reads target model state and routes through a dual memory system - discrete retrieval for known patterns, a neural policy for novel situations. The strategy bank grows with every run. We walk through a controlled evaluation of a frontier language model as one case study among broader deployments. The core finding: behavioral gaps are distributed. Cognitive bias chaining, consensus loops, creative framing, credential-based extraction, and role-inversion attacks all succeed independently, yet no single strategy dominates - patching the top vector leaves most of the surface exposed. Successful multi-turn attacks oscillate between refusal and compliance, revealing that per-turn enforcement is stateless. The same search framework applies to any customer-defined policy or behavioral expectation - business requirements, regulatory compliance, tone guidelines, over-refusal, domain-specific constraints. Discovered patterns feed into a production guardrail cascade; production bypasses re-enter the search loop, so evaluation and enforcement co-improve over time.

Bio: Anish Das Sarma is the Founder and CEO of Reinforce Labs, a startup dedicated to ensuring responsible adoption of enterprise AI systems. A repeat founder, Anish previously built Trooly, an AI-powered identity and trust platform that was acquired by Airbnb. At Airbnb, he went on to lead key initiatives in AI, trust & safety. Most recently, Anish served as a Director of Engineering at Google, where he led large-scale AI/ML teams across Google Ads Safety. Anish holds a Ph.D. in Computer Science from Stanford University and a B.Tech in Computer Science from IIT Bombay. Across more than a decade in AI, he has combined deep technical expertise with hands-on operational leadership, building products and teams across multiple domains of applied ML.

Shutong  Zhang
Software Engineer,  Youtube Trust and Safety, Meta 

Speaker

Title: Tool-MCoT: Tool Augmented Multimodal Chain-of-Thought for Content Safety Moderation

Abstract: The growth of online platforms and user content requires strong moderation systems that can handle complex inputs from various media types. While large language models (LLMs) are effective, their high computational cost and latency present significant challenges for scalable deployment. To address this, we introduce Tool-MCoT, a small language model (SLM) fine-tuned for content safety moderation leveraging external framework. By training our model on tool-augmented chain-of-thought data generated by LLM, we demonstrate that the SLM can learn to effectively utilize these tools to improve its reasoning and decision-making. Our experiments show that the fine-tuned SLM achieves significant performance gains. Furthermore, we show that the model can learn to use these tools selectively, achieving a balance between moderation accuracy and inference efficiency by calling tools only when necessary.

Bio: Shutong Zhang is a software engineer on the YouTube Trust and Safety team, working on text and image understanding. He received his master's degree from Stanford University, where his research mainly focused on computer vision and multimodal LLMs.

Ankitesh Gupta  
Senior Software Engineer, Trust and Safety, Meta 

Title: Behavioral Graph Enrichment: Using LLMs to Bridge Structural Gaps in Integrity Networks

Abstract: Graph propagation algorithms are the workhorses of large-scale abuse detection—expanding a small set of confirmed malicious seeds into comprehensive enforcement queues. But on sparse integrity graphs, where adversaries deliberately minimize their structural footprint, these methods fail: seeds sit in isolated neighborhoods while the bulk of abuse remains undiscovered. We present a behavioral graph enrichment method that uses a multi-agent LLM pipeline as a graph preprocessing step—scoring entities on coordinated-abuse dimensions and creating weighted edges between structurally disconnected but behaviorally similar entities. On a real-world Facebook Groups inauthenticity graph, adding a small set of behavioral edges improves candidate precision by 7–16 percentage points across four standard algorithms, without modifying any downstream logic.

Bio: Ankitesh is a Senior Software Engineer at Meta, working within Facebook Trust and Safety. His work focuses on understanding and measuring the extent of harmful content and abuse across the platform—building AI-powered systems that maximize the effectiveness of limited human review budgets to enable accurate, scalable integrity measurement. He operates at the intersection of machine learning and human judgment, driving efforts that inform how Meta quantifies and responds to platform-wide risks.