• Akto’s GenAI Security Testing solution pinpoints vulnerabilities in the security protocols responsible for purifying the output of LLMs.
  • The solution has numerous features, incorporating over 60 test cases that address diverse aspects of GenAI vulnerabilities, including prompt injection, excessive reliance on particular data sources, and more.

Akto Io Inc., an application programming interface company, has recently introduced GenAI Security Testing, a novel solution designed to bolster the security of generative artificial intelligence and large language models.

Positioned as the inaugural proactive platform for generative AI security testing globally, this innovative service is tailored to alleviate mounting apprehensions regarding vulnerabilities linked to the implementation of AI and related services.

As a result of the need for more automated, efficient workflows, Akto contends that the use of AI and LLMs in business has surged, bringing with it new risks like denial-of-service attacks, unauthorized prompt injections, and inaccurate data from relying too much on AI outputs without due diligence. It is now necessary for security teams to find a new, automated method to secure LLMs at scale, as hackers are becoming more inventive in their exploits of LLMs.

Akto’s GenAI Security Testing solution identifies security vulnerabilities in the measures employed to cleanse the output of LLMs. This service is adept at recognizing efforts to inject malicious code for remote execution, cross-site scripting, and other potential attacks that might enable attackers to retrieve session tokens and system information. Akto also assesses the susceptibility of LLMs to produce inaccurate or irrelevant reports.

Comprising numerous features, the solution includes over 60 test cases that address diverse aspects of GenAI vulnerabilities, encompassing prompt injection, excessive dependence on specific data sources, and additional factors. Akto’s experts in generative AI security have meticulously crafted these test cases to guarantee the utmost level of protection for organizations implementing generative AI models.

Ankush Jain, Chief Technology Officer at Akto, said, “The tests try to exploit LLM vulnerabilities through different encoding methods, separators and markers. This specially detects weak security practices where developers encode the input or put special markers around the input.”

Akto is a startup supported by venture capital, having secured USD 4.5 million in funding, as reported by Tracxn. Noteworthy investors in the company include Accel Partners Inc., Akshay Kothari (Co-founder and Chief Operating Officer of Notion Labs Inc.), Renaud Deraison (Co-founder of Tenable Inc.), and Milin Desai (Chief Executive of Sentry Inc.).