Confident Security, ‘the Signal For Ai,’ Comes Out Of Stealth With $4.2m

As consumers, businesses, and governments flock to nan committedness of cheap, fast, and seemingly magical AI tools, 1 mobility keeps getting successful nan way: How do I support my information private?

Tech giants for illustration OpenAI, Anthropic, xAI, Google, and others are softly scooping up and retaining personification information to amended their models aliases show for information and security, moreover successful immoderate endeavor contexts wherever companies presume their accusation is disconnected limits. For highly regulated industries aliases companies building connected nan frontier, that grey area could beryllium a dealbreaker. Fears astir wherever information goes, who tin spot it, and really it mightiness beryllium utilized are slowing AI take successful sectors for illustration healthcare, finance, and government. 

Enter San Francisco-based startup Confident Security, which intends to beryllium “the Signal for AI.” The company’s product, CONFSEC, is an end-to-end encryption instrumentality that wraps astir foundational models, guaranteeing that prompts and metadata can’t beryllium stored, seen, aliases utilized for AI training, moreover by nan exemplary supplier aliases immoderate 3rd party.

“The 2nd that you springiness up your information to personification else, you’ve fundamentally reduced your privacy,” Jonathan Mortensen, laminitis and CEO of Confident Security, told TechCrunch. “And our product’s extremity is to region that trade-off.”

Confident Security came retired of stealth connected Thursday pinch $4.2 cardinal successful seed backing from Decibel, South Park Commons, Ex Ante, and Swyx, TechCrunch has exclusively learned. The institution wants to service arsenic an intermediary vendor betwixt AI vendors and their customers – for illustration hyperscalers, governments, and enterprises.

Even AI companies could spot nan worth successful offering Confident Security’s instrumentality to endeavor clients arsenic a measurement to unlock that market, said Mortensen. He added that CONFSEC is besides well-suited for caller AI browsers hitting nan market, for illustration Perplexity’s precocious released Comet, to springiness customers guarantees that their delicate information isn’t being stored connected a server location that nan institution aliases bad actors could access, aliases that their work-related prompts aren’t being utilized to “train AI to do your job.”

CONFSEC is modeled aft Apple’s Private Cloud Compute (PCC) architecture, which Mortensen says “is 10x amended than thing retired location successful position of guaranteeing that Apple cannot spot your data” erstwhile it runs definite AI tasks securely successful nan cloud.

Like Apple’s PCC, Confident Security’s strategy useful by first anonymizing information by encrypting and routing it done services for illustration Cloudflare aliases Fastly, truthful servers ne'er spot nan original root aliases content. Next, it uses precocious encryption that only allows decryption nether strict conditions.

“So you tin opportunity you’re only allowed to decrypt this if you are not going to log nan data, and you’re not going to usage it for training, and you’re not going to fto anyone spot it,” Mortensen said. 

Finally, nan package moving nan AI conclusion is publically logged and unfastened to reappraisal truthful that experts tin verify its guarantees. 

“Confident Security is up of nan curve successful recognizing that nan early of AI depends connected spot built into nan infrastructure itself,” Jess Leao, partner astatine Decibel, said successful a statement. “Without solutions for illustration this, galore enterprises simply can’t move guardant pinch AI.”

It’s still early days for nan year-old company, but Mortensen said CONFSEC has been tested, externally audited, and is production-ready. The squad is successful talks pinch banks, browsers, and hunt engines, among different imaginable clients, to adhd CONFSEC to their infrastructure stacks. 

“You bring nan AI, we bring nan privacy,” said Mortensen.