SafetyKit

SafetyKit's AI agents do the dirty work for big tech and finance companies so humans don't have to.

Today companies rely on armies of low-paid outsourced humans to protect their platforms from fraud, harassment, and crime. It’s super important work, but the jobs are difficult and sometimes traumatizing, and it’s impossible to keep up. Our mission is to end this manual toil.

We learned this domain first-hand at Stripe and Airbnb, where first we built software for human reviewers, and then built models to replace them.

We’re a group of engineers from Stripe, Meta, Netflix, MIT, and Princeton that love language models and unironically love B2B SaaS. We know the economy is going to be rearranged over the next 4 years and we want to be in the driver’s seat.

We build vertical AI that works and have some of the world’s largest companies as paying customers. Patreon, Eventbrite, Upwork, Character.ai, Substack, Faire, along with a $10B+ marketplace, a $100B+ payments company, and more rely on SafetyKit in production.

High-Level Overview

SafetyKit is an AI-driven platform that builds multimodal AI agents specialized in risk, compliance, and safety for marketplaces, payment platforms, and fintech companies. Its technology automates the detection and response to fraud, prohibited activities, scams, and other compliance risks across various content types including text, images, financial transactions, and product listings. By leveraging advanced OpenAI models like GPT-5, SafetyKit delivers scalable, precise, and real-time risk management solutions that reduce manual review workloads and improve accuracy. The platform serves a broad range of customers from startups to Fortune 500 companies, protecting hundreds of millions of end users and enabling faster, more confident decision-making in risk operations[1][3][5].

For an investment firm perspective, SafetyKit’s mission is to end manual toil in risk and compliance reviews by replacing human reviewers with AI agents, thereby scaling review capacity and enhancing accuracy without increasing headcount. Its investment philosophy centers on leveraging cutting-edge AI to transform labor-intensive compliance processes. Key sectors include fintech, marketplaces, payments, and online platforms facing complex regulatory and safety challenges. SafetyKit’s impact on the startup ecosystem is significant, as it enables rapid product scaling and geographic expansion without compromising safety or compliance, fostering innovation in high-risk digital commerce environments[2][5].

From a portfolio company perspective, SafetyKit builds AI agents that automate risk detection and compliance enforcement. It serves risk teams at large tech companies, fintechs, and marketplaces who struggle with manual, costly, and error-prone review processes. The problem it solves is the inefficiency and human cost of manual risk reviews, enabling customers to respond swiftly to emerging threats like fraud, scams, illegal products, and child exploitation. Growth momentum is strong, with daily processing of over 16 billion tokens (up from 200 million six months prior), expansion into new risk domains, and adoption by major clients including Patreon, Eventbrite, Upwork, and large payment platforms[1][3][5].

Origin Story

SafetyKit was founded in 2023 by David Graunke and Steven Guichard, engineers with deep experience in risk and compliance software at Stripe and Airbnb. The founders witnessed firsthand the challenges of manual review teams—difficult, traumatizing work that could not keep pace with evolving threats. They initially built software tools to assist human reviewers and then pivoted to develop AI models to replace them. The founding team includes engineers from Stripe, Meta, Netflix, MIT, and Princeton, reflecting a strong technical pedigree focused on language models and B2B SaaS[5].

The idea emerged from the founders’ direct experience with the inefficiencies of manual risk reviews and the opportunity to apply advanced AI to automate these processes. Early traction came from deploying AI agents that could handle complex policy tasks and multimodal content analysis, gaining paying customers among startups and Fortune 500 companies. A pivotal moment was raising $27 million in Series A funding in 2025 to accelerate platform development and scale integrations with existing risk infrastructures[2][5][6].

Core Differentiators

• Multimodal AI Agents: SafetyKit’s agents analyze text, images, financial transactions, and product listings, enabling nuanced detection of fraud, scams, illegal content, and compliance violations beyond legacy keyword or rule-based systems[1][3].

• Model Matching Architecture: Each agent is tailored to a specific risk category and uses the optimal OpenAI model (e.g., GPT-4.1, GPT-5) for that task, improving accuracy and efficiency in content review[1].

• Real-Time, Scalable Enforcement: The platform processes billions of tokens daily with 24/7 global policy coverage, enabling rapid, automated decision-making that scales with customer growth[1][3].

• Deep Integration with Risk Workflows: SafetyKit automates complex policy tasks such as merchant onboarding, transaction laundering detection, anti-money laundering (AML), and child safety, reducing manual review burdens and operational costs[1][4].

• Strong Developer and Customer Ecosystem: Trusted by major marketplaces, fintechs, and payment platforms, SafetyKit supports multiple languages and compliance regimes, facilitating global expansion for clients[3][5].

• Continuous Learning and Adaptation: The AI learns from every decision to catch sophisticated fraud patterns and evolving threats that traditional systems miss[4].

Role in the Broader Tech Landscape

SafetyKit rides the trend of AI-driven automation in risk, compliance, and safety, a domain historically reliant on manual labor and rule-based systems. The timing is critical as digital commerce, marketplaces, and fintech platforms face increasing regulatory scrutiny, rising fraud sophistication, and scaling challenges that outpace human review capacity. Advances in large language models and multimodal AI enable SafetyKit to close the gap between scale and nuance, providing real-time, context-aware enforcement that legacy solutions cannot match[1][4].

Market forces favor SafetyKit’s growth: the explosion of online transactions, the rise of AI-generated content, and the need for faster onboarding and fraud detection all drive demand for automated, intelligent risk management. By enabling platforms to maintain safety without slowing growth, SafetyKit influences the broader ecosystem by setting new standards for compliance automation and helping platforms capture high-intent traffic safely in an AI-driven economy[1][4].

Quick Take & Future Outlook

Looking ahead, SafetyKit is poised to expand its AI agents into new risk domains and deepen integrations with major payment and marketplace infrastructures. The company will likely continue leveraging the latest OpenAI models to enhance accuracy and speed, while scaling its global policy coverage and language support. Trends such as AI-generated commerce, evolving regulatory environments, and increasing demand for real-time risk decisions will shape SafetyKit’s journey.

Its influence will grow as more platforms adopt AI-powered compliance tools to reduce operational burdens and human error, enabling safer digital economies. SafetyKit’s mission to end manual toil in risk review aligns with a broader shift toward vertical AI solutions that transform complex workflows, positioning it as a key player in the future of compliance automation and platform safety[1][2][4][5].