|
Enterprises are embracing generative AI (Gen AI) to streamline operations and increase revenue by letting AI perform manual, tedious tasks. However, the potential for data loss hinders many deployments. Organizations are concerned about their key intellectual property and sensitive data being inadvertently disclosed. GenAI-based apps were a top data loss vector, with 43% of enterprises having experienced a data loss event via a GenAI-based application. |
企业正在采用生成式人工智能（Gen AI），让人工智能执行繁琐的手动任务，从而简化运营并增加收入。但是，数据丢失的可能性阻碍了许多部署。各组织担心其关键知识产权和敏感数据会被无意中泄露。基于Genai的应用程序是最大的数据丢失载体，有43％的企业经历过通过基于Genai的应用程序发生过数据丢失事件。 |
|
Existing data loss prevention (DLP) tools are based heavily on regular expression (regex) logic that functions well with known, structured data types like personally identifiable information (PII). However, the regex approach does not lend itself to GenAI applications where intellectual property and other unstructured data are commonplace. Conventional solutions being redeployed for GenAI applications can be a burden to configure and administer and also tend to generate significant false positive alert noise that burdens security teams. GenAI applications are a different beast that benefits from a new DLP approach. |
现有的数据丢失防护 (DLP) 工具在很大程度上基于正则表达式 (regex) 逻辑，该逻辑可以很好地处理已知的结构化数据类型，例如个人身份信息 (PII)。但是，正则表达式方法不适用于知识产权和其他非结构化数据司空见惯的GenAI应用程序。为GenAI应用程序重新部署传统解决方案可能会给配置和管理带来负担，而且往往会产生严重的误报警音，给安全团队带来负担。GenAI 应用程序与众不同，它受益于新的 DLP 方法。 |
|
GenAI applications are also being rapidly adopted, frequently without oversight from security teams. It is easy for employees to fire up their browsers and visit ChatGPT, Claude, or Perplexity to complete their tasks. Such tools may be unauthorized or risky AI applications, and security teams frequently lack visibility and control over AI apps. |
GenAI 应用程序也被迅速采用，通常不受安全团队的监督。员工很容易启动浏览器并访问 ChatGPT、Claude 或 Perplexity 来完成任务。此类工具可能是未经授权或存在风险的人工智能应用程序，安全团队通常对人工智能应用程序缺乏可见性和控制力。 |
|
Any solution needs to overcome the traditional DLP challenges of minimizing administrative overhead and reducing alert noise. While existing solutions might solve that challenge for existing data loss vectors like email and endpoints, GenAI applications pose different risks to sensitive information. Enterprises need to have an adequate inventory of AI assets, identify and assess shadow AI, enforce AI policies, and continuously guide end users to avoid inadvertent data leakage. GenAI is different in that solutions need to prevent leakage of unstructured sensitive data like intellectual property and source code. Compliance requirements mean that solutions also need to detect personally identifiable information (PII) and cardholder information affected by Payment Card Industry Data Security Standard (PCI-DSS) mandates. |
任何解决方案都需要克服传统 DLP 挑战，即最大限度地减少管理开销和减少警报噪音。尽管现有解决方案可以解决电子邮件和端点等现有数据丢失载体的挑战，但GenAI应用程序对敏感信息构成了不同的风险。企业需要有足够的人工智能资产清单，识别和评估影子人工智能，执行人工智能政策，并持续指导最终用户避免无意中数据泄露。GenAI的不同之处在于，解决方案需要防止知识产权和源代码等非结构化敏感数据的泄漏。合规要求意味着解决方案还需要检测受支付卡行业数据安全标准 (PCI-DSS) 要求影响的个人身份信息 (PII) 和持卡人信息。 |