Menlo Security targets real-time AI agent security with MARS platform
As AI agents gain access to enterprise systems and sensitive data, security teams need greater visibility into their actions. Real-time monitoring and policy enforcement are becoming essential as a part of prompt injection technology. With these challenges in mind, Menlo Security Inc. has sought to respond with Menlo Agent Runtime Security (MARS). In this new […] The post Menlo Security targets…
Real-time monitoring and policy enforcement are becoming vital in the era of AI agents accessing enterprise systems and sensitive data. Menlo Security Inc. has introduced Menlo Agent Runtime Security (MARS) to tackle these challenges. Menlo Security's Chief Product Officer, Ramin Farassat, highlights the risks associated with AI agents, including data exfiltration and potential manipulation through prompt injection attacks.
He explained that AI agents, which often blindly process input data, can be deceived by malicious commands hidden within seemingly legitimate requests. Farassat emphasized that while 100% protection against prompt injection remains elusive, implementing it as one of several defensive measures is crucial. Menlo Security's MARS operates by setting policies and controlling agent actions, even when conflicting with other commands.
Additionally, it scrutinizes the data agents consume in real-time, masking sensitive information to mitigate risks.
Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
