MIND Joins Anthropic Cyber Verification Program to Advance Data Loss Prevention
MIND, a provider of a platform that enforces guardrails on what types of data can be shared with a large language model, this week revealed it has been accepted into the Anthropic Cyber Verification Program.
Company CEO Eran Barak said MIND is now the first provider of a data loss prevention (DLP) platform to be included in the Anthropic Cyber Verification Program. MIND will add Anthropic models to the AI platform it developed to discover sensitive data that is being shared via a prompt. Ultimately, the goal is to develop more advanced data classification methods, model threats and identify data exfiltration patterns indicative of insider threat behavior to provide an autonomous DLP capability.
Anthropic originally launched this program to provide defenders, researchers, and data security companies with access to Claude models, including Opus 4.7 and the Mythos model that is currently only being made available to a limited number of organizations because of its ability to discover vulnerabilities in code.
The Anthropic verification process requires a cybersecurity vendor to demonstrate its commitment to ethical security practices and prove that its use cases align with defensive security objectives. The program is designed to maintain strict boundaries to ensure AI models are used only for legitimate security purposes.
It’s not clear yet how many organizations are concerned about how much sensitive data is being shared with AI models. While there have been several incidents where sensitive data surfaced in the output of a prompt made to a large language model (LLM), there have not been enough incidents yet to discourage organizations from using AI models. In the absence of major incidents, it becomes more challenging to convince business leaders to allocate funding.
However, that may only be a question of time given how long it takes for sensitive data to be potentially used to train the next iteration of an AI model.
Regardless of the number of cybersecurity incidents, sharing sensitive data with an LLM likely violates any number of compliance mandates that eventually auditors will discover. In most cases, the fines and penalties levied will then wipe out the productivity gains that might have been gained.
Most recently, MIND extended the reach of its data loss prevention (DLP) platform to artificial intelligence (AI) agents that are increasingly being deployed across organizations with little to no regard for cybersecurity implications. The fundamental challenge AI agents create is they can autonomously create, access, transform and share data at machine speed. The probability those AI agents are going to share sensitive data beyond the four walls of the proverbial organization at this juncture is high. They also present cybercriminals with a tempting target that, if compromised, provides a means to potentially access massive amounts of data.
In an ideal world, cybersecurity teams should be proactively addressing these issues but the fact is end users are adopting AI faster than most cybersecurity teams are able to effectively track. Unfortunately, few of those end users fully appreciate all the security implications involved as AI becomes more pervasively employed.
