As artificial intelligence technologies become more and more mainstream in the corporate world, many companies face the issue of keeping confidential information secure. Expedient, a Pittsburgh-based cloud infrastructure company, is aiming to address that issue by providing companies with what they call a "Secure AI Gateway."
"As more and more users embrace [AI] in their everyday work, the risk of leaking sensitive or confidential information goes up exponentially, which represents a material risk for an organization," Expedient CEO Bryan Smith said in a prepared statement. "Secure AI Gateway gives organizations a foundation for using AI in a safe, transparent and scalable way."
Expedient will provide companies with access to several leading AI models, including OpenAI's GPT-4, Meta's Llama, Anthropic's Claude and others. The gateway controls what data and models users have access to and utilizes a single-sign-on authentication for all users, which allows administrators to monitor what employees are using the models for.
"The Secure AI Gateway is your nexus point so that you can choose how you want to interact with AI," Bradley Reynolds, SVP of AI at Expedient, said. "There are a bunch of folks working on solid, private foundational models. We will provide access to all of those, and the vetting process is 'is this something of high quality."
In addition to the chat functions, custom applications can be built on top of the gateway by request.
Artificial intelligence models are rapidly developing and sometimes changes to these complex models are for the worse. Last weekend, GPT-4 users began experiencing issues utilizing the service to write and debug code, issues that were not present previously. To address issues like this, Reynolds said that Expedient will offer users the ability to backtrack to earlier versions of the models.
"We can host a private, dedicated model for you so that you are in control of the software," Reynolds said. "With Meta, they have Llama 2 and Llama 3. Let's say that you were using Llama 2 but decided that you want all of the things in Llama version three. We can download Llama 3 and give it to you and if you decide that that's not good you can revert back to the previous version."
Public perception of AI is mixed. A Pew Research Study found that of Americans, 18% are more excited than afraid of AI, 45% are equally excited as they are concerned and 37% are more afraid than excited. The number one reason cited for being more afraid than excited was "loss of human jobs."
"It's pretty rare, maybe even never happened before, that a technology affects every job potentially and that's disconcerting," Reynolds said. "I think that the best companies will be getting ahead of it with their employees and having policies and talking to them and training them to find ways to increase their output and protect their jobs as opposed to looking at it as 'how do we cut a bunch of costs.' I think that the winners long term will figure out how to accelerate their businesses, not just trim expense lines based on some technology."
Expedient is not the only Pittsburgh based company to debut technology seeking to secure AI in the workplace. Preamble, a startup based out of downtown, launched a similar service last month.
