Noma is building tools to spot security issues with AI apps

Companies are concerned that their eagerness to adopt AI has made them more vulnerable to cyberthreats. Per a recent poll of over 350 IT leaders, more than half of the executives surveyed said the complexity of AI applications weakened their organization’s cybersecurity posture. More than two-fifths of executives believe their security teams lack the skills needed to protect their AI applications and workloads, per another survey.

The demand for solutions to secure AI apps has spawned an entire cohort of startups. HiddenLayer and Protect AI, for example, focus on defending AI systems from adversarial attacks, while Cranium provides visibility into AI systems at the application level.

A new venture emerging from stealth today, Noma Security, does all of this and more, claims co-founder and CEO Niv Braun. The startup is developing tools to identify vulnerable data pipelines and code in data science environments, as well as threats like prompt injection attacks.

Braun launched Noma with Alon Tron after leaving customer experience automation firm Verint, where he noticed a “significant blind spot” for security teams in the data and AI lifecycle.

“Existing tools are designed for the traditional software lifecycle,” Braun said. “However, the data and AI lifecycle is inherently different. It involves distinct R&D processes, relies on different technologies, and faces unique technical vulnerabilities within AI models.”

Noma is designed to help spot — and fix — misconfigurations in the components used to develop and run AI applications. For instance, the platform scans for “sensitive” data in model training datasets, such as text containing personally identifiable information.

Braun argues that companies can be better equipped to deal with threats if they can consolidate AI app security functions in one dashboard that can be deployed in cloud-based and self-hosted environments.

“Many single-function security solutions are flooding the market, trying to capitalize on the AI hype,” he said. “By helping companies understand their AI application footprint and giving them governance controls, Noma is not only helping organizations minimize and mitigate risk, but also giving them the confidence they need to expand their AI usage faster and to more high-value parts of the business.”

Noma, founded in 2023, already has paying customers, Braun says, including Fortune 500 clients in B2B software, financial services and retail. The startup recently closed a $25 million Series A funding round that Braun said was preempted by investors; Noma had cash to spare.

“We decided to accept the [$25 million], even with substantial funds still in the bank, to accelerate our development and especially our go-to-market teams and efforts,” he said, adding that Noma plans to double the size of its 20-person team in the coming year.

Ballistic Ventures led Noma’s Series A, which brings the company’s total capital raised to $32 million. Noma raised $7 million in a previously-undisclosed seed round led by Glilot Capital Partners with participation from Cyber Club London.