SentinelOne Inc. and Snyk Ltd. today introduced new tools for securing artificial intelligence agents.

NYSE-traded SentinelOne provides a cybersecurity platform called Singularity that helps companies protect software containers, employee devices and other assets. Snyk, in turn, offers tools that developers use to fix vulnerabilities in their code. The company is backed by more than $1.3 billion in funding.

SentinelOne’s first new tool, Prompt AI Agent Security, can automatically identify the MCP servers used by a company’s AI agents. It then applies cybersecurity guardrails to those servers to prevent insecure use. The tool can prevent AI agents from sending business data to a third-party application. It also mitigates other risks, such as situations where an AI agent gains access to an internal system that it’s not authorized to use.

Prompt AI Agent Security is rolling out alongside a second new tool called Prompt AI Red Teaming. According to SentinelOne, it finds vulnerabilities in AI applications by simulating cyberattacks. The software can simulate malicious prompts, attempts to compromise a model’s training data and other threats.

SentinelOne’s Singularity platform finds cybersecurity risks by analyzing telemetry from organizations’ infrastructure. Going forward, its Singularity AI SIEM module will ingest that telemetry using an “AI-native data pipeline.” The feature is based on technology that SentinelOne obtained last year through the $225 million acquisition of Observo AI.

The data pipeline organizes telemetry before uploading it to Singularity. It enriches the data with external information, normalizes it and filters unnecessary details. SentinelOne says that the feature can reduce data noise, or unnecessary telemetry, by up to 80%, which lowers the infrastructure costs associated with processing it.

Cybersecurity platforms often install a lightweight program, or agent, on the systems they’re used to protect. The agent collects telemetry and sends it to a cloud-based backend for processing. Such data transfer isn’t possible in air-gapped environments, sensitive corporate networks that are isolated from the web. That limitation makes deploying cybersecurity software a challenge. 

As part of today’s update, SentinelOne debuted an AI workload security tool that is specifically designed for air-gapped environments. Prompt Security On-Premises stores the telemetry that it collects on customers’ own infrastructure instead of sending it to the cloud. It uses SentinelOne’s new AI data pipeline to filter unnecessary telemetry before performing analysis.

“For too long, organizations in highly regulated sectors have faced a trade off between the speed of AI security and total data sovereignty, privacy and control – especially for airgapped networks,” said Ana Pinczuk, president of product and technology at SentinelOne.

The AI security tools that Snyk debuted today address many of the same use cases as SentinelOne’s new features. The main highlight is a technology called Snyk Evo AI-SPM. It automatically creates an inventory of the language models, MCP servers and other AI assets in a company’s code repositories. It then scans those assets for cybersecurity risks.

Evo AI-SPM powers three AI agents that are currently in preview. The first ensures that the third-party AI components used by a company’s developers are secure. According to Snyk, the other agents help developers test their AI applications for vulnerabilities and block AI risks such as prompt responses that contain training dataset excerpts. 

“Agentic architectures turn governance into a software supply chain problem,” said Snyk Chief Innovation Officer Manoj Nair. “Our value is confirming which findings are real and exploitable, using ground truth data from a decade of enterprise deployment that no AI model can produce alone. Claude finds. Snyk confirms. The agent fixes only what’s real.”

Photo: SentinelOne