[Digital Today reporter Chi-gyu Hwang] SentinelOne and Snyk have each unveiled new tools for AI agent security, SiliconANGLE reported on March 24 (local time).

SentinelOne launched "Prompt AI Agent Security". It automatically identifies MCP servers used by enterprise AI agents and applies security guidelines to block use in risky situations. It prevents AI agents from sending corporate data to external applications and blocks unauthorised access to internal systems.

"Prompt AI Red Teaming", unveiled alongside it, is a tool for finding vulnerabilities in AI applications. It detects security weaknesses by simulating various cyberattacks, including malicious prompt simulations and attempts to compromise model training data.

The AI security information and event management (SIEM) module in SentinelOne's Singularity platform collects telemetry through an "AI native data pipeline". Telemetry is operational data automatically collected from systems, devices and networks, including logs, events and performance metrics.

SentinelOne said it secured the technology through a startup it acquired last year for $225 million. The company said it reduces unnecessary telemetry by up to 80 percent, cutting data processing costs.

SentinelOne also released a tool for air-gapped network environments that are completely cut off from the internet. "Prompt Security On-Premise" stores collected telemetry on customers' own infrastructure instead of sending it to the cloud.

Ana Pinczuk (아나 핀추크), president of products and technology at SentinelOne, said, "Companies in highly regulated sectors have long been forced to choose between AI security speed and full control of their data." She added, "Now this problem can be solved even in air-gapped networks."

Snyk unveiled "Evo AI-SPM", which automatically inventories AI assets such as language models and MCP servers in corporate code repositories and scans for security risks. Evo AI-SPM is used for three agents currently in preview.

The agents are responsible respectively for checking the security of third-party AI components used by developers, testing AI application vulnerabilities and blocking AI risks such as training data leaks.

Manoj Nair (마노즈 나이르), chief innovation officer at Snyk, said, "Agent architecture turns governance into a software supply chain issue." He added, "The company's value lies in verifying which vulnerabilities are real threats using real data accumulated over 10 years in enterprise environments."