EP287: Creating Trust at Global Scale with Local AI: Reken-ing with Shuman Ghosemajumder

This section covers Schumann Ghosemajumder’s insights from his early days at Google, focusing on how the environment was ripe for innovation in trust and safety, especially around advertising and fraud prevention. Schumann shares stories about launching Gmail with a GigaByte of storage to attract users and how this opened avenues for misuse by cybercriminals. He narrates the development of automation tools used by bad actors, starting with logic-based automation and evolving into AI-driven methods. The discussion emphasizes how cybercriminals target major enterprises first, then commoditize their methods for broader use, illustrating the efficiency of their resource allocation and adaptation to new technologies. The concept of cybercriminals operating like R&D teams is highlighted, with a focus on the pattern of targeting high-value targets first to develop sophisticated tools before reducing the scale of attacks to smaller targets.

Schumann recounts the origin of the term 'credential stuffing,' a label that he and his team helped popularize during his time at Shape Security. Initially a little-known attack method involving automated testing of stolen username and password pairs across multiple sites, it became the leading security threat by 2019. The segment discusses how cybercriminals leverage large breaches and the commonality of password reuse among users. The narrative underscores the importance of industry awareness and collective effort in understanding and combating this threat, illustrating how simple automation, when scaled, can pose significant risks.

This section introduces Schumann’s new startup, Reckon, which aims to rebuild trust in digital communications. Schumann underscores the decline of trust due to rampant AI-generated scams, deepfakes, and automated fraud—an AI DDoS—paralyzing communication channels like email and phone systems. Reckon’s approach is to create a trust layer that employs local, on-device AI models to analyze communication data without compromising privacy. By doing so, Reckon seeks to defend against the scalable, AI-driven deception tactics that threaten the integrity of online interactions. Schumann emphasizes the importance of strong network effects and bidirectional device protection, offering a vision for a more secure, trustworthy internet infrastructure.

In the closing parts of the podcast, Schumann shares practical advice for listeners on how to spot malicious content and stay secure in an era of widespread AI manipulation. He advocates for a healthy degree of paranoia—constantly verifying information and sources—as a key security mindset. Schumann also discusses the importance of combining human awareness with technological solutions, especially local AI, to combat deepfakes, scams, and other forms of AI-assisted fraud. The episode concludes with reflections on building a new trust layer for the internet, balancing privacy with security, and fostering strong network effects to make digital interactions safer.