Wajo AI Agent Bets on Privacy to Rival Meta Muse and OpenAI Dot
Backed by Vinod Khosla and former Google engineers, the Fo agent uses virtual credit cards and human fallbacks to handle daily tasks securely.
A new personal artificial intelligence agent called Fo, built by startup Wajo, is attempting to carve out market share against Meta’s Muse and OpenAI’s Dot by centering its architecture around data privacy and user trust. The platform has secured backing from venture capitalist Vinod Khosla, alongside former Google executives Jeff Dean and Gokul Rajaram, though the company has not disclosed the size of its funding round.
The personal-agent sector has accelerated rapidly in recent weeks, with multiple startups launching tools designed to autonomously handle scheduling, purchasing, and communications on behalf of users. Wajo’s entry into this crowded field relies on a specific thesis: consumers will hesitate to hand comprehensive access to their personal lives, messages, and financial details to corporations whose primary revenue models depend on advertising or broad data utilization.
Khosla argued that platforms operated by companies entrenched in the ad business inherently compromise user privacy, positioning Wajo as an alternative architected specifically for safety. This distinction matters because personal agents require deep system-level permissions to function effectively, creating unprecedented privacy risks compared to standard chatbots or search tools.
Fo operates across iMessage, WhatsApp, and a dedicated web interface where tasks are organized into categories such as scheduling, gifting, and life administration. The agent can initiate phone calls to businesses or individuals to book or cancel services. During early testing, the system successfully contacted an airline to check for flight upgrades. However, when tasked with calling a personal contact to deliver a reminder, the interaction encountered friction, particularly regarding how clearly the AI disclosed its non-human identity to the recipient.
As autonomous voice capabilities expand across the industry, the protocols governing how an AI identifies itself during outbound calls remain largely unresolved. Competitors like Instinct and Muse are also introducing calling features gradually, signaling that regulatory and social norms around synthetic voice disclosure will become a defining battleground for the sector.
To address security vulnerabilities inherent in automated commerce, Wajo developed a virtual credit card system that allows Fo to execute payments without exposing the user’s actual financial credentials. The company has filed a patent application for this technology. The agent also maintains its own email address to correspond directly with businesses and individuals. When websites or digital services actively block AI traffic, Fo alerts the user to intervene manually or establishes a verification group connecting the user, the agent, and the business.
One structural differentiator for Wajo is its ability to route tasks to human workers when an automated process fails or hits a technical limitation, blending AI autonomy with gig-economy fallback mechanisms.
Wajo was founded by Shivani Poddar, an engineer who previously worked at Meta, Google, and DeepMind. Her background includes contributing to the development of Facebook’s first assistant and working on safety protocols for Gemini. Poddar noted that while large technology companies possess immense technical depth, they sometimes struggle to translate that capability into grounded consumer products. She emphasized that building trust requires architectural decisions made before launch, particularly because current machine learning frameworks lack reliable methods for forcing an AI agent to permanently unlearn or delete personal data once it has been processed.
To mitigate unwanted intrusions, Wajo implemented a trusted contact list that governs which individuals the agent is permitted to call. Recipients who receive calls from Fo can instruct the system never to contact them again, a preference the platform records and enforces.
Since its launch in late September, Wajo reported a tenfold increase in usage and confirmed the service is now active in 107 countries. Specific user counts were not provided. The company stated it plans to introduce multi-user social functionalities in the coming months, aiming to allow different users’ agents to interact while maintaining strict privacy boundaries.
