Become a member of our daily and weekly newsletters for the latest updates and exclusive content about leading AI coverage. Leather
Enterprise employees would like to benefit from AI tools -whether their employer likes it or not. This not -approved use, which is known as Shadow AI, increases dramatically: as much as as if 96% of the work Doing employees with AI through non-business accounts. Whether it is done unintentionally or malicious, this can leak the very sensitive and own data of a company.
Security platform Cyberhaven Says that it can solve this problem by keeping track of data line or data lifecycli with different users and end points. The company has specific large origin models (LLIMS) for this task and today announces Linea AI, the next generation of its platform, intended to help stop shadow AI and to predict which marked incidents can be the most dangerous.
“It manifests itself in this form of origin: you understand where data comes from, which have access to it, in all different endpoints, with all your users,” Nishhant Doshi, the most important product and development officer in Cyberhaven, told Venturebeat in an exclusive interview.
90% Reduction of incidents that require manual assessment
According to Cyberhaven’s analysis of the workflows of 3 million employees, AI use grew 485% Between March 2023 and March 2024. Employees are increasingly sharing sensitive data: almost 83% of legal documents and approximately 50% of the source code, research and development materials and HR and employee reports that share employees with AI go to non-business AI accounts.
In order to prevent this non -sanctioned use and to protect sensitive company data, Linea AI uses an LLIM that is trained on billions of actual data flows for business results. Equipped with computer vision and multimodal AI, the data analysis of images, screenshots, technical diagrams and other materials can. A new position “Let Linea AI decide” now assesses autonomous policy violations and the severity of the incident to help reduce the Security Operations Center (SOC) alert fatigue.
“So just like the Great Language Model (LLM) that predicts the next word, we predict what the following actions will be,” Doshi explained.
Cyberhaven claims that as a result of this, customers see a 90% reduction in incidents that require manual assessment and a decrease in the average time of 80% to respond (MTTR) on security incidents with regard to data security. The company’s tools can discover more than 50 critical risks per month that are not detected by traditional tools.
“Cyberhaven shows us exactly how our data moves and is used throughout the organization, which means we have visibility that is not found with traditional security tools,” said Prabathhat Karanth, CSO and CIO of Family Financial App Greenlight. “Now we have one platform that not only relates to traditional data loss (DLP) and insider risk management, but even understands how people use data throughout our organization.”
Doshi explained that, while traditional approaches have focused on pattern agreement – identifying network and data patterns to detect anomalies and vulnerabilities – Cyberhaven performs content and context inspection. The platform investigates data and offers context around on the basis of Lineage tracks.
“So if you download something, you send it to me, I send it to another five people, they send it to another five people – that’s origin,” Doshi explained.
How Cyberhaven protects the most valuable data from Enterprises with AI
The range of Cyberhaven is powered by Frontier AI models and a Neural Network Architecture transformer. It uses a multi-phase Retrieval-Augmented Generation (RAG) engine to refine its LLIM to analyze the most valuable data of a company and “go to the needle in the haystack,” said Doshi.
The platform performs intelligent screenshot analysis, which has been a “persistent blind spot” in data security, said Aaron Arkeen, senior security engineer in earned wage access platform DailyPay.
So, for example, say that a security team wants to prevent screenshots from leaving the company. There may be thousands, and they have to mix up to determine if it is a harmless cat meme or a screenshot with product schedules.
“It is difficult to detect the exfiltration of engineering designs, AI models, research data, product trout maps, let alone prevent,” said Arkeen.
Keep an eye on users
Cyberhaven is now taking cyber security one step further than detection with its new autonomous, AI-driven Let Linea decisions function that searched through data and user logbooks to help security teams understand the severity of the incident. The platform understands screenshots, PDFs, source code and other digital materials and can offer context based on data line, explains Doshi. Subsequently, it can distinguish whether a specific incident should be viewed by human analysts.
“We try to predict the following action based on all the historical knowledge that we have: this is an abnormal event, or this is a benign event,” said Doshi. “We mention that concept, because you really look at the data and understand that data in a deeply.”
Arkeen explained that when it comes to insider risks, security teams perform improved monitoring to create information flows about specific users that are marked as an increased risk (based on any number of factors).
“Let’s say I made you improve, you were busy today, 150 events were generated,” he said. “I would have to go through each of those manually,” this is things as usual. “” This looks a bit suspicious. ” “This one looks really suspicious.” And after that I have others to continue.
For example, the platform has been able to detect users who send data to their personal OneDrive accounts or synchronize sensitive files with iCloud, Doshi said. A malignant step further is that employees leave a company and try to bring sensitive data.
“In real time we can prevent users or a set of user uploading sensitive data to these public LLMs,” said Doshi. “We can warn them and also train them” when they do something unintentionally or naive.
DailyPay, for his part, has been able to reduce MTTR by 65% because Linea offers a digestible AI summary, said Arkeen. Typical tools for data loss prevention (DLP) require many staff sources to get that kind of visibility.
He looked at other DLP providers, including Netskope, DTEX systems and next DLP, but eventually settled at Cyberhaven because of the strategy for data line. It was different from everything he had seen in the industry, he said.
“It saves us a lot of time on escalation and triadaging and also prevention,” said Arkeen. “Linea AI consistently identifies nuanced risks that traditional systems will absolutely miss.”
Source link
