网络安全
Iboss 发布 AI DLP用于实时数据控制跨云和AI
在 Google 上的首选来源中添加 Unite.AI
iboss 宣布 AI DLP于2026年11月3日，是一款数据损失预防产品，旨在实时检查AI工具，云应用和其他在线目的地的内容。 公司计划在2026年12月开设私人预览。
该产品针对的是传统数据损失预防方面的限制。 基于正则表达式和模式匹配对预先定义格式的搜索的规则，如一个帐号或ID字符串，而依赖标签的控件可能会错过被错误标签或从未标签的文件。 例子包括共享到云驱动器的文件，提交给网络应用程序的表格和粘贴到AI工具的源代码，所有这些文件可以携带敏感材料而无需安装固定模式。
内容检查和政策控制
AI DLP将AI和机器学习分类与深度内容检查相结合，在数据移动时对其进行检查。 文件分析和光学字符识别将检查范围扩大到图像内部的文件和文字，包括贴上标签、错误标签或无标签的内容。 设计以一个文件或即时文件的实质内容作为强制执行的依据，包括格式规则和元数据没有提供可靠信号时。
政策可以由用户、团体或部门确定、标记或停止涉及源代码、客户数据、个人可识别信息和知识产权的转让。 iboss DLP文档还描述了来源和目的地范围界定，命名的例外，以及包括屏蔽，文件捕捉，记录，提醒和严重性分配在内的响应。 违反规定的行为可进入有风险分数的事件，并附有事件时间表和被捕获的文件证据，使分析人员可以进行调查和结案，而不只是个别记录条目。
加密交通和Agent活动
内容意识控制要求能见度超过目的地。 iboss对AI DLP 称的解释是，基于网络的AI工具一般需要加密的会话检查这一级别的控制；没有它，系统可能会看到数据正在向何处发展，但不会看到即时发送或发送的文件。 这种能见度允许内容政策以与其他网络上传相同的方式适用于AI聊天。
控制范围也超越了将文本粘贴到聊天窗口的人。 它们可以管理自动agent的AI服务、工具和API使用和保留agent活动的记录。 这将agent治理与用于交互式AI会话的同一执行模式联系起来，涵盖软件和员工发起的数据移动。
标签、审计证据和平台架构
在存在Microsoft信息保护敏感性标签的地方，AI DLP读取并强制其内含。 产品文档描述了每个标签，每个目的地允许或屏蔽使用Microsoft Purview标签同步的决定，而内容检查则继续覆盖没有标签的文件。 因此，标签与内容分析一起成为一项政策投入，从而减少对每个载有正确元数据的文件的依赖。
报告包括GDPR、HIPAA、PCI-DSS和SOX的数据损失风险评分和一击报告，目的是向审计员提供发现和停止的证据。 iboss 称每位客户的流量在专用容器中处理，并配有孤立的SSL密钥和专用IP地址，客户数据不共混。 AI DLP与该公司的安全网关、CASB、GenAI保护和零信任网络接入服务并列
2026年11月5日晚12点ET提前举行30分钟在线演示。 实际区别是宣布设计与可用性：iboss为人类和agent驱动的数据运动设置了单一的内容感知控制平面，私人预览是将这一设计付诸实践的预定步骤。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Cybersecurity
Iboss Unveils AI DLP for Real-Time Data Controls Across Cloud and AI
Add Unite.AI to your preferred sources on Google
iboss announced AI DLP on November 3, 2026, a data-loss prevention product designed to inspect content in real time across AI tools, cloud applications and other online destinations. The company plans to open a private preview in December 2026.
The product targets limitations in conventional data-loss prevention. Rules based on regular expressions and pattern matching look for formats defined in advance, such as an account number or ID string, while label-dependent controls can miss files that were labeled incorrectly or never labeled. Examples include a file shared to a cloud drive, a form submitted to a web application and source code pasted into an AI tool, all of which can carry sensitive material without fitting a fixed pattern.
Content Inspection and Policy Controls
AI DLP combines AI and machine-learning classification with deep content inspection to examine data as it moves. File analysis and optical character recognition extend that inspection to documents and text inside images, covering content that is labeled, mislabeled or unlabeled. The design bases enforcement on the substance of a file or prompt, including when format rules and metadata provide no reliable signal.
Policies can be scoped by user, group or department to identify, flag or stop transfers involving source code, customer data, personally identifiable information and intellectual property. The iboss DLP documentation also describes source and destination scoping, named exceptions, and responses including blocking, file capture, logging, alerts and severity assignment. Violations can roll into risk-scored incidents with event timelines and captured-file evidence, giving analysts a case to investigate and close instead of only individual log entries.
Encrypted Traffic and Agent Activity
Content-aware controls require visibility into more than a destination. iboss’s explanation of AI DLP says that web-based AI tools generally require encrypted-session inspection for this level of control; without it, a system may see where data is going but not the prompt or file being sent. That visibility allows content policies to apply to AI chats in the same way as other web uploads.
The controls also extend beyond a person pasting text into a chat window. They can govern which AI services, tools and APIs an autonomous agent may use and retain a record of agent activity. This connects agent governance to the same enforcement model used for interactive AI sessions, covering data movement initiated by software as well as by employees.
Labels, Audit Evidence and Platform Architecture
Where Microsoft Information Protection sensitivity labels exist, AI DLP reads and enforces them inline. The product documentation describes per-label, per-destination allow or block decisions using Microsoft Purview label synchronization, while content inspection continues to cover unlabeled documents. Labels therefore become one policy input alongside content analysis, reducing dependence on every document carrying the correct metadata.
Reporting includes data-loss risk scoring and one-click reports for GDPR, HIPAA, PCI-DSS and SOX, intended to give auditors evidence of what was found and stopped. iboss says each customer’s traffic is processed in dedicated containers with isolated SSL keys and dedicated IP addresses, and that customer data is not co-mingled. AI DLP sits alongside the company’s Secure Web Gateway, CASB, GenAI Protection and Zero Trust Network Access services, with a shared console, policy engine and data lake.
A 30-minute online demonstration is scheduled for November 5, 2026, at 12:00 p.m. ET, ahead of the December preview. The practical distinction is between announced design and availability: iboss has laid out a single content-aware control plane for human and agent-driven data movement, and the private preview is the scheduled step toward putting that design into practice.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.