Artificial intelligence safety platform startup WitnessAI Inc. today announced the launch of Agentic Security, which extends its “confidence layer” for enterprise artificial intelligence beyond large language models to cover autonomous and semi-autonomous AI agents.

The new capability is designed to give security teams visibility, control and protection as organizations increasingly deploy agents that can access internal systems, external application programming interfaces and third-party tools.

The Agentic Security capability seeks to address the issue whereby employees and engineering teams are deploying agents that accept natural language inputs and execute backend tasks through tool calls. The problem is that security teams lack visibility into which agents exist, which tools they access and how to protect them from adversarial manipulation.

WitnessAI argues that agents introduce a fundamentally different attack surface than traditional AI chat interfaces. Because agents can act automatically on every permission they are granted, a single malicious or manipulated prompt can cascade through tool calls, API requests and database queries before a human notices, creating risks including prompt injection, data exfiltration and unauthorized actions that conventional network and endpoint security tools were not built to detect.

Agentic Security is aimed at addressing the security gap by automatically discovering agentic activity across high-risk environments such as Claude Desktop and plugins, ChatGPT with enabled plugins, VS Code with AI extensions and locally running agent frameworks including LangChain, LlamaIndex, CrewAI and AutoGPT.

The platform can distinguish standard chat sessions from agentic sessions by analyzing tool usage in traffic payloads and maps which public and private Model Context Protocol servers are being accessed to enrich them with intent and functional classifications.

The platform also focuses on accountability by linking agent actions back to the humans who initiated them.

Agentic Security connects human and agent identities and captures execution context at runtime to allow security teams to see who triggered a workflow and what the agent attempted to do, even when agents communicate with other agents. The ability to find attribution allows for a single policy framework and audit trail across both human and agentic workforces.

Other features of the new offering include bidirectional runtime defense, with prompts scanned before execution to detect and block prompt injection, jailbreaking and role-playing attacks, while sensitive data such as credentials and personally identifiable information is tokenized in real time. Agent outputs are also inspected before delivery to enforce policy compliance and filter harmful content, with enforcement based on behavioral intent rather than simple keyword matching.

“AI workflows are maturing and starting to cross corporate and cloud LLMs, bots and agents,” said Chief Executive Rick Caccia. “We are the only AI security vendor that can secure every AI interaction, everywhere, with a unified solution. The alternative is trying to stitch together secure workflows using network proxies, firewalls, data loss protection products and extended detection and response agents. In short, the alternative is a complex mess.”

WitnessAI is a venture capital-backed startup that has raised a single round of $27.5 million in May 2024. Investors in the company include Google Ventures (GV Management Co.) and Ballistic Ventures.

Image: WitnessAI

Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.

15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more
11.4k+ theCUBE alumni — Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network.

About SiliconANGLE Media

SiliconANGLE Media is a recognized leader in digital media innovation, uniting breakthrough technology, strategic insights and real-time audience engagement. As the parent company of SiliconANGLE, theCUBE Network, theCUBE Research, CUBE365, theCUBE AI and theCUBE SuperStudios — with flagship locations in Silicon Valley and the New York Stock Exchange — SiliconANGLE Media operates at the intersection of media, technology and AI.

Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.