Confident Security Raises $4.2M to Advance AI-Powered Security

The Growing Concern of Data Privacy in the Age of AI

With the increasing adoption of artificial intelligence by individuals, organizations, and governmental bodies, a critical question arises: How can data privacy be maintained?

Major technology companies, including OpenAI, Anthropic, xAI, and Google, are collecting and storing user data to refine their AI models and enhance security measures. This practice occurs even within enterprise settings where data confidentiality is expected. For industries subject to strict regulation, or those pioneering new technologies, this ambiguity can present a significant obstacle. Apprehensions regarding data destinations, access permissions, and potential usage are hindering AI implementation in sectors like healthcare, finance, and public administration.

Introducing Confident Security: A Privacy-Focused Solution

San Francisco-based startup Confident Security positions itself as a solution, aiming to be a leading provider of AI data privacy. Their product, CONFSEC, is an end-to-end encryption tool designed to operate with foundational AI models.

CONFSEC ensures that prompts and associated metadata are protected from storage, unauthorized viewing, and utilization for AI training purposes – even by the model provider or any external parties.

“The moment data is transferred to another entity, privacy is compromised,” explains Jonathan Mortensen, founder and CEO of Confident Security, in an interview with TechCrunch. “Our product is engineered to eliminate this necessary trade-off.”

Seed Funding and Market Positioning

Confident Security emerged from stealth mode on Thursday, announcing $4.2 million in seed funding secured from Decibel, South Park Commons, Ex Ante, and Swyx. The company intends to function as an intermediary between AI vendors and their clientele, encompassing hyperscalers, governmental organizations, and large enterprises.

Mortensen suggests that even AI companies could benefit from integrating CONFSEC into their offerings to access the enterprise market. He further notes that the tool is particularly suitable for emerging AI browsers, such as Perplexity’s Comet, providing users with assurances that their sensitive information won’t be stored insecurely or used to replicate their work.

Inspired by Apple’s Private Cloud Compute

CONFSEC’s architecture is modeled after Apple’s Private Cloud Compute (PCC), which Mortensen describes as “significantly superior to existing solutions in guaranteeing Apple’s inability to access user data” during secure cloud-based AI tasks.

Similar to Apple’s PCC, Confident Security’s system initially anonymizes data through encryption and routing via services like Cloudflare or Fastly. This prevents servers from directly accessing the original data source or content.

Subsequently, it employs advanced encryption techniques that restrict decryption to specific, controlled conditions.

“Conditions can be set to prohibit data logging, prevent use for training, and restrict access to unauthorized individuals,” Mortensen clarifies.

Finally, the software executing the AI inference is publicly logged and available for scrutiny, allowing experts to validate its security guarantees.

Industry Validation and Future Outlook

“Confident Security is proactively addressing the critical need for trust-based infrastructure in the future of AI,” states Jess Leão, partner at Decibel. “Solutions like this are essential for enabling enterprise adoption of AI.”

Although the company is relatively new, having been founded a year ago, Mortensen confirms that CONFSEC has undergone rigorous testing, external audits, and is now ready for production deployment. The team is currently engaged in discussions with banks, browser developers, and search engine providers to integrate CONFSEC into their respective infrastructure stacks.

“Clients provide the AI; we deliver the privacy,” concludes Mortensen.