Protect AI’s new Sightline provides insight into artificial intelligence vulnerabilities

Protect AI Inc., an artificial intelligence and machine learning cybersecurity company, today announced Sightline, a vulnerability database that provides insights into known and emerging AI and machine learning vulnerabilities and early warnings to protect against threats. announced the system.

Sightline is designed to help organizations rapidly detect, assess, and remediate AI risks associated with their AI ecosystems and environments, enabling enterprises to adopt AI and machine learning models with confidence and security. .

The rise of AI and machine learning has been well documented, but supply chain risks in the libraries, packages, frameworks, underlying models, and datasets used in AI have received more attention. That was just recently. Protect AI recognizes that while supply chains are at the root of many AI innovations and digital transformations, existing vulnerability databases do not adequately cover AI or provide the necessary details about AI risks, and AI applications and claims that it does not address all the complexities of securing machine learning. model. This is the sight line step.

“Traditional vulnerability databases are inadequate, and Sightline uniquely addresses this gap,” explains Ian Swanson, CEO of Protect AI. “Sightline does more than just add noise to vulnerability databases. It focuses precisely on AI and ML vulnerabilities and is backed by Hunter, a hacker community specializing in AI/ML.”

Protect AI's “huntr” bug bounty community is said to be the industry's only AI and machine learning hacker community, with over 15,000 experts identifying under-researched areas and vulnerabilities in fundamental AI components. are discovering. Sightline leverages insights uncovered by huntr to help enterprises ensure the security of their AI applications.

The Hunter bug bounty community currently discovers and fixes over 50 unique vulnerabilities in the AI/machine learning supply chain every month. Notable previous discoveries include remote code execution issues in Hugging Face Transformers, Nvidia Corp.'s Triton Interface Server, and PyTorch Serve.

Sightline also includes penetration testing tools that can quickly assess vulnerabilities within your AI environment and provide additional context about the impact of issues. Sightline integrates with Protect AI's Radar security posture management product to provide a contextual view of vulnerabilities within a customer's AI supply chain. This enables fast and accurate responses to potential threats by automatically assessing the relevance of vulnerabilities within the unique context of an AI application.

Sightline is building on the growth of Protect AI, which includes Guardian, a secure gateway launched in January that helps organizations apply security policies to machine learning models to prevent malicious code from entering the environment. Join our continuing suite of AI security tools.

Image: Protecting AI