Confident security, “a reference for artificial intelligence”, comes out of the ghost with $ 4.2 million

Since consumers, companies and governments are on the promise of cheap and fast artificial intelligence tools, one question continues on the way: How can I maintain the privacy of my data?

Technology giants such as Openai, Anthropic, Xai, Google and others collect user data and retain calmness to improve their models or safety and security monitoring, even in some contexts of institutions in which companies assume their information outside the borders. For industries or companies with the high organization they adopt on the border, this gray area can be deals. Fears about the place where the data is going, who can see it, and how it can be used to slow the adoption of artificial intelligence in sectors such as health care, financing and government.

Enter the start starting in San Francisco The confident securityWhich aims to be a “signal for artificial intelligence”. The company, Confsec, is a comprehensive encryption tool that exceeds the founding models, ensuring that the demands and descriptive data cannot be stored, viewed or used to train artificial intelligence, even by the model provider or any third party.

“The second that you give up from your data for another person, you have mainly reduced your privacy,” Jonathan Mortensen, founder and CEO of Confident Security, told Techcrunch. “The goal of our product is to remove this comparison.”

The confident security of the ghost came out on Thursday with $ 4.2 million from the seed financing from Despell, South Park Comong, former, and Swix, I learned Techcrunch exclusively. The company wants to be a mediator seller among the sellers of artificial intelligence and their customers – such as Easter, governments and institutions.

Mortensen said that artificial intelligence companies can see the value in providing a confident security tool for institutions as a way to cancel the lock of this market. He added that Confsec is also suitable for artificial browsers GuiltyTo give customers to ensure that their sensitive data is not stored on a server that the company or poor actors can access, or that their claims related to work are not used to “train artificial intelligence to do your work.”

Confsec is designed similar to the Apple (PCC) structure, which Mortensen says, “10x is better than anything in terms of ensuring that Apple cannot see your data” when you manage some of the tasks of artificial intelligence safely in the cloud.

Like PCC from Apple, the Customent Security system works by giving the first character to unknown data by encrypting and directing it through services such as Cloudflare or Fastly, so you never see servers or original content. Next, advanced encryption that only allows the encryption to decomposition under strict conditions.

“So you can say that you are only allowed to decipher this if you do not record the data, you will not use it for training, and you will not allow anyone to see it,” said Mortensen.

Finally, the program that manages the inference of artificial intelligence publicly is registered for review so that experts can check its guarantees.

“The confident security progresses on the curve in the realization that the future of artificial intelligence depends on the confidence in the same infrastructure,” Jess Liao, a partner in Despell, said in a statement. “Without solutions like this, many institutions cannot simply proceed with artificial intelligence.”

It is still early for the year -old company, but Mortensen said that Confsec was tested, and it was audited externally, and it is ready for production. The team holds talks with banks, browsers and search engines, among other potential customers, to add Confsec to their infrastructure chimneys.

“You are preparing artificial intelligence, we bring privacy,” said Mortensen.