Confident Security, ‘the Signal for AI,’ comes out of stealth with $4.2M

In a world increasingly reliant on AI technologies, the concern for data privacy looms large. As businesses, consumers, and governments eagerly embrace affordable and speedy AI solutions, the pressing question remains: How can personal and sensitive data be safeguarded? Major tech firms such as OpenAI, Anthropic, xAI, and Google have been gathering user data to enhance their models and ensure safety. However, in regulated sectors like healthcare and finance, this practice raises significant concerns about data usage and accessibility. These apprehensions are contributing to a slowdown in AI implementation across several critical industries. To tackle these issues, San Francisco-based startup Confident Security is stepping into the spotlight as a potential game-changer in data privacy. The company has introduced CONFSEC, its innovative end-to-end encryption tool designed to envelop foundational AI models, ensuring that prompts and metadata remain confidential and inaccessible for AI training, including from third-party providers. Jonathan Mortensen, the founder and CEO of Confident Security, emphasized the crucial need for privacy in AI interactions. "The moment you share your data, you compromise your privacy," he stated. Mortensen aims to eliminate that dilemma by providing a solution that protects user data. On Thursday, Confident Security emerged from stealth mode, announcing it has secured $4.2 million in seed funding from investors including Decibel, South Park Commons, Ex Ante, and Swyx. The startup aspires to act as a crucial intermediary between AI providers and their clients, such as enterprises and government entities. Mortensen noted that even AI companies could benefit from integrating CONFSEC into their services to better serve enterprise customers. The tool is also positioned to enhance privacy for new AI browsers, such as Perplexity’s recently launched Comet, by ensuring that sensitive data is not stored where it could be accessed by unauthorized parties or used to train AI models inappropriately. Inspired by Apple’s Private Cloud Compute architecture, CONFSEC offers a robust level of data protection. According to Mortensen, this approach is significantly more secure, allowing operations that limit data visibility while still performing essential AI tasks in the cloud. The methodology involves anonymizing data through encryption and routing via services like Cloudflare or Fastly, preventing the original content from being exposed. Moreover, it employs advanced encryption techniques that restrict decryption to specific conditions, ensuring that the data is not logged or used for unauthorized purposes. The software's operations are transparent, enabling experts to review the processes to verify compliance with its privacy guarantees. Jess Leão, a partner at Decibel, commented on the importance of Confident Security's innovations, stating, "Confident Security is ahead of the curve in recognizing that the future of AI depends on trust built into the infrastructure itself. Without solutions like this, many enterprises simply can’t move forward with AI." While Confident Security is still in its early stages, Mortensen confirmed that CONFSEC has undergone testing and external audits, making it ready for production. The team is actively engaging with banks, browsers, and search engines to integrate CONFSEC into their operational frameworks. "You bring the AI, we bring the privacy," Mortensen concluded.