AI models security startup Cranium spun out of KPMG Studio

SECURITY

Cranium.ai Corp., a startup looking to ensure trust and security in artificial intelligence models, today launched as a standalone company from KPMG Studio and raised venture capital funding from SYN Ventures Management L.P.

The company is the first startup to come out of KPMG Studio, KPMG International Ltd.’s tech incubator. It helps startup founders test the viability of their idea in the market and implement their approach in real time.

Cranium provides an AI security and trust platform that operates by mapping AI pipelines to validate their security and monitor for adversarial threats. The platform detects and prevents AI attacks, such as data poisoning, model inversion, model evasion and backdoor detection or membership interferences.

The technology integrates with existing environments without interrupting the way organizations work to test, train and deploy their AI models. Cranium’s experts have also developed a playbook that security teams can use alongside the software to protect their AI systems and adhere to existing U.S. and EU regulatory standards.

Key features of the platform include collaboration between data science and cybersecurity teams: to ensure that key AI lifecycle stakeholders have a common operating picture across teams to improve visibility and collaboration. Inventory AI pipelines capture in-development and deployed AI pipelines, including all associated assets throughout the AI lifecycle.

Cranium’s platform establishes an AI security framework that provides security and data science teams with a foundation for building a proactive and holistic AI security program. The platform also captures and quantifies AI security risk and establishes continuous monitoring.

“Awareness and adoption of AI solutions across the enterprise is at a fever pitch, with generative AI models showcasing the ‘magic’ of AI to the general public,” said Jonathan Dambrot, founder and chief executive officer of Cranium. “In tandem with this acceleration, major regulatory bodies, such NIST, European Commission and the FDA, are moving swiftly to set guidelines for organizations deploying AI systems.”

Dambrot added that “the inherent nature of AI models learning from new inputs creates an entirely unique set of attack methods that require novel security techniques to account for these threats.” Cranium, he said, brings visibility, trust and a new level of security to cybersecurity and data science teams when it comes to the entire AI ecosystem of an organization.

Image: Cranium