Agentic AI identity verification startup Vouched gets $17M to build trust in autonomous agents – News

An identity verification startup called Vouched Identity Inc. is turning its attention away from humans to focus on artificial intelligence agents after raising $17 million in a Series A round of funding today from Spring Rock Ventures.

The startup has made a name for itself as a provider of digital identity verification tools, with its software used to scan individual’s selfie photos and compare them with supplied documentation, such as passports and driving licenses, to verify that people are whom they say they are.

But what sets it apart from other identity verification providers is that it’s now laser-focused on identifying AI agents. As these autonomous software systems built using generative AI are increasingly taking over the web, completing tasks, booking holidays and so on for human users, businesses have an urgent need to be able to identify them and make sure they can be trusted.

This is exactly what Vouched is trying to do with its “Know Your Agent” product, which is a suite of tools for verifying the identities of AI agents and monitoring their behavior. It also operates an extensive database of AI agents called “Know That Agent,” which allows businesses to check the reputation of each agent that visits their site.

The startup believes it’s providing a much-needed service for regulated enterprises in industries such as e-commerce and digital marketplaces. With its platform, companies can identify the AI agents that access their services and continuously monitor their behavior, with specific modules for checking their credentials, level of risk and continuous compliance.

The Know Your Agent suite offers two main tools. The first, Agent Shield, was developed to help businesses identify any AI agents accessing their websites instead of humans. It identifies every session where an AI agent is acting on behalf of a human, helping site operators to understand how much agentic traffic they’re experiencing. Each time an AI agent visits a website, it creates a session ID for that agent, so that companies can determine what kinds of actions they’re taking on their sites.

Vouched Chief Executive Peter Horadan told News that no other company can offer this kind of capability, to tell AI agents apart from human users. “There is no established mechanism, so we created a proprietary heuristic, which works much like spam detection,” he explained. “Both we and some other companies in the industry have proposed new protocols for the identification of agents, but it is so early that none of these have been adopted yet.”

The second tool is called Agent Bouncer, which supports open frameworks such as the Model Context Protocol to advance interoperability and trust in agent ecosystems. The idea here is to protect against malicious activity.

When an AI agent connects with an MCP server or gateway so it can take actions on a particular website, Agent Bouncer will identify what that agent is, decide if it’s trustworthy, determine who it is acting for, and then ensure that the individual has given the agent explicit permission to take those actions.

Horadon said one of the major risks associated with AI agents is that they can aid in so-called “phishing” attacks, where cybercriminals attempt to steal users’ passwords and other private information. Moreover, he said some company employees provide their work login credentials to AI agents willingly, in order to make their jobs easier, without considering whether or not that agent can be trusted.

“This is an astonishing breach of security protocols,” he claimed. “If you are an employee and you give up your work credentials to a third party [without authorization], you can and probably should be fired.”

Another risk for businesses is that their customers may also give up their login credentials to AI agents, resulting in fraudulent orders that look as if they have been made by legitimate human users. Horadon said Vouched can prevent that from happening by blocking these malicious agents from taking actions on their websites. And it can also help protect corporate secrets, he said, pointing to the complete lack of privacy in agent interactions.

“There is no privacy with anything you give to an AI agent,” Horadon said. “The New York Times has an active subpoena against OpenAI and can retrieve any chat, confidential or not. If your employees have given up their credentials to ChatGPT, that’s the same as sending your prospect list, customer list or any other confidential information to the New York Times.”

Because of these risks, Horadan says, any company that’s looking to use AI agents needs to be sure that it can trust them, which is where its identification and verification tools come in useful. “We are doubling down on KYA’s vision of trustworthy autonomous agents and scaling our platform to match the pace of a rapidly evolving digital world,” he said.

With the funding from today’s round, Vouched intends to build on the capabilities of its KYA suite, grow its KTA database and collaborate with more industry partners to help ensure the safety of future agent-driven interactions.

Image: News/Dreamina AI