This Startup Secretly Detects Fraud For Fortune 500s
Y Combinator Startup PodcastFull Title
This Startup Secretly Detects Fraud For Fortune 500s
Summary
Variants, a company specializing in AI agents for risk and compliance, has emerged from stealth, announcing a $21 million Series A funding round.
The company provides automated solutions for fraud, identity, and content reviews, serving major corporations and marketplaces by leveraging advanced AI to combat increasingly sophisticated threats.
Key Points
- Variants operates in stealth mode for three years to protect sensitive data and avoid alerting bad actors to their detection methods, positioning themselves as a "secret weapon" for their clients.
- The company's AI agents are used to automate critical reviews for platforms like GoFundMe, identifying fraudulent fundraisers by analyzing user behavior, identity, and fundraiser details.
- Variants also handles complex identity verification for marketplaces and gig economy platforms, as well as Know Your Business (KYB) verifications to ensure individuals are legitimately linked to the businesses they represent, including identifying complex ownership structures and potential risks like sanctioned countries or money laundering.
- A significant technical challenge for Variants was aggregating and processing vast amounts of data from disparate sources, including data hidden behind user interfaces, which required developing agents capable of interacting with existing human-facing dashboards.
- The evolution of AI, particularly LLMs, has enabled Variants to move beyond a patchwork of deterministic systems and human analysts, creating a self-healing, dynamic fraud detection system that can adapt rapidly to new threats.
- During elections, Variants' AI agents detected sophisticated fraud rings, including state-sponsored actors pushing narratives, by analyzing the context of entities and their relationships within large datasets, a feat difficult for isolated human analysis.
- The company's small but highly capable engineering team, amplified by AI coding agents, achieves output comparable to a much larger team, allowing for rapid feature development and autonomous processing of customer requests.
- The founders' origin story stems from their experience on Apple's fraud engineering team, where they identified inefficiencies in traditional fraud detection methods and were driven by a sense of duty to solve these problems with advanced technology.
- Securing their first enterprise customer, IAC, was a lengthy but crucial step, validating their approach to automating content review for marketing materials that were previously reliant on slow and costly human moderation.
- A serious accident involving the CEO highlighted the need to scale the company beyond relying solely on the founders, emphasizing the importance of robust processes and team structure.
Conclusion
Variants demonstrates the power of AI agents in tackling complex risk and compliance challenges that were previously unautomatable, offering a significant advantage over traditional methods.
The company's stealthy approach and focus on deep technical solutions, coupled with the rapid advancements in AI, position them as a critical infrastructure provider for the modern digital economy.
The founders' origin story and resilience in the face of adversity underscore a strong sense of purpose and a commitment to solving significant industry problems.
Discussion Topics
- How do companies balance the need for transparency with the strategic advantage of operating in stealth mode?
- What are the ethical considerations when AI agents are used to detect potentially harmful content and activities online?
- As AI technology rapidly advances, what is the future role of human analysts in risk and compliance roles?
Key Terms
- AI agents
- Software programs that can perform tasks autonomously, often mimicking human cognitive functions like reasoning and decision-making.
- Fortune 500
- A list of the 500 largest companies in the United States by revenue.
- Stealth mode
- A business strategy where a company operates without public disclosure of its product, technology, or business plans.
- Series A funding
- The first major round of venture capital financing for a startup after seed funding.
- Content review
- The process of examining text, images, or videos for compliance, policy violations, or fraud.
- Fraud reviews
- Analyzing transactions, applications, or behaviors to identify and prevent fraudulent activities.
- Identity reviews
- Verifying the authenticity and legitimacy of an individual or entity's identity.
- UBO verifications
- Ultimate Beneficial Owner verifications, which identify the individuals who ultimately own or control a company.
- KYC (Know Your Customer)
- A set of procedures used by businesses to verify the identity of their clients.
- KYB (Know Your Business)
- A process used by businesses to verify the identity and legitimacy of other businesses they are engaging with.
- LLMs (Large Language Models)
- Advanced AI models trained on vast amounts of text data, capable of understanding and generating human-like text.
- Deterministic systems
- Systems that produce the same output for the same input, often relying on fixed rules.
- Reverse ETL
- The process of moving data from a data warehouse back into operational systems for use in business processes.
- PII (Personally Identifiable Information)
- Information that can be used to identify a specific individual.
Timeline
Variants is building purpose-built AI agents for risk and compliance, automating content review, fraud reviews, and identity reviews at scale for large companies and marketplaces.
The company operated in stealth for three years due to the sensitive nature of the data and issues they handle, aiming to avoid alerting adversaries to their detection methods.
Variants uses AI agents to verify GoFundMe fundraisers, such as distinguishing legitimate causes from fraudulent ones, like multiple people claiming to be related to a deceased individual.
Variants also performs complex identity reviews for marketplaces and gig economy platforms, verifying individuals' identities against selfies and driver's licenses, and conducts KYB verifications for businesses.
The core technical challenge for Variants was the data problem, involving the aggregation and reasoning over petabytes of data from diverse and sometimes unstructured sources, including data accessed through user interfaces.
The evolution of AI, particularly LLMs and AI agents, has allowed Variants to move beyond deterministic systems and human analysts, creating a self-healing and rapidly evolving fraud detection system.
Variants' AI agents detected complex fraud rings, including state-sponsored actors during elections, by analyzing relationships between entities and contextualizing data, which would be difficult for traditional methods.
Variants has a small team of five software engineers who, amplified by AI coding agents, achieve a high output comparable to a much larger team.
The founders, Kareem and Michael, met as fraud engineers at Apple and were driven to start Variants to solve the inefficiencies they observed in enterprise fraud detection.
Their first customer was IAC, a large publicly traded company, for whom they automated the review of marketing content, a process previously handled by a large team of human agents.
The CEO, Kareem, was hit by a truck, highlighting the company's initial reliance on founders and the need to scale operations.
Episode Details
- Podcast
- Y Combinator Startup Podcast
- Episode
- This Startup Secretly Detects Fraud For Fortune 500s
- Official Link
- https://www.ycombinator.com/
- Published
- March 31, 2026