As companies more and more depend on cloud-driven platforms and AI-powered instruments to speed up digital transformation, the stakes for safeguarding delicate enterprise information have reached unprecedented ranges. The Zscaler ThreatLabz 2025 Knowledge@Danger Report reveals how evolving know-how landscapes are amplifying vulnerabilities, highlighting the crucial want for a proactive and unified strategy to information safety.
Drawing on insights from greater than 1.2 billion blocked transactions recorded by the Zscaler Zero Belief Alternate between February and December 2024, this 12 months’s report paints a transparent image of the info safety challenges that enterprises face. From the rise of information leakage by generative AI instruments to the undiminished dangers stemming from e mail, SaaS functions, and file-sharing providers, the findings are each eye-opening and pressing.
The 2025 Knowledge@Danger Report sheds mild on the multifaceted information safety dangers enterprises face in at this time’s digitally enabled world. Among the most noteworthy traits embrace:
- AI apps are a serious information loss vector: AI instruments like ChatGPT and Microsoft Copilot contributed to thousands and thousands of information loss incidents in 2024, significantly social safety numbers.
- SaaS information loss is surging: Spanning 3,000+ SaaS apps, enterprises noticed greater than 872 million information loss violations.
- E-mail stays a number one supply of information loss: Practically 104 million transactions leaked billions of situations of delicate information.
- File-sharing information loss spikes: Among the many hottest file-sharing apps, 212 million transactions noticed information loss incidents.
AI functions: A brand new information loss hotspot
Generative AI instruments reminiscent of ChatGPT and Microsoft Copilot are revolutionizing how enterprises work—however not with out penalties. These platforms accounted for 4.2 million information loss violations, revealing how private identifiers, mental property, and monetary information are routinely in danger.
SaaS ecosystems: Simplifying workflows, complicating safety
Greater than 872 million information loss incidents have been flagged throughout SaaS platforms. Well-liked functions reminiscent of Microsoft 365, Salesforce, and Google Workspace, which have the biggest share of violations, spotlight the strain between collaboration and compliance.
E-mail: A legacy danger with perennial penalties
Regardless of newer instruments and platforms, e mail stays on the forefront of information loss. Microsoft Alternate and Gmail collectively noticed 104 million transactions containing billions of information loss incidents. The commonest leaks included medical information, social safety numbers, and supply code.
File-sharing platforms: Productiveness with a heaping aspect of danger
File-sharing giants like Google Drive, Microsoft OneDrive, and Dropbox logged 212 million transactions that concerned information loss. Delicate data—starting from proprietary supply code to monetary data—flowed unchecked in billions of particular person violations throughout these transactions.
Whereas the report reveals large volumes of information loss throughout the preferred functions, it additionally offers a roadmap for organizations to behave decisively earlier than information leaks or exfiltration occur. By adopting a unified, AI-driven strategy to information safety, companies can flip these dangers into alternatives and safe information throughout each channel, wherever it resides.
Finest observe suggestions from the 2025 Knowledge@Danger Report embrace:
- Use AI to find and classify your information: Implement a Zero Belief Structure (ZTA), enabling superior information loss prevention (DLP) insurance policies throughout endpoints and networks, and leveraging AI-powered platforms to establish dangers in real-time. By taking these steps, enterprises can safeguard their information whereas enabling productiveness and innovation to thrive.
- Perceive your information loss channels: Map out all of the channels by which information flows inside and out of doors your group—e mail, SaaS apps, AI instruments (e.g., Microsoft Copilot), BYOD, cloud storage, and bodily storage gadgets. Every channel presents distinctive dangers and requires tailor-made safety controls.
- Lean in your Zero Belief Structure: Transition from a perimeter-based safety mannequin to a ZTA that enforces least-privileged entry. Use identity-based entry management, granular insurance policies, and Safe Entry Service Edge (SASE) to examine all web site visitors, section networks, and reduce your group’s assault floor.
- Safe GenAI and AI instruments with granular controls: For generative AI instruments like ChatGPT and Microsoft Copilot, implement granular controls on person periods, reminiscent of enter or output restrictions. Block unsafe prompts that may expose delicate information throughout person interactions. Moreover, monitor anomalies in person conduct (e.g., extreme queries) and flag or block actions that violate information safety insurance policies.
As enterprise AI transforms workflows and accelerates innovation, the challenges of managing and securing information develop in parallel. From delicate prompts leaked in generative AI instruments to information loss throughout SaaS platforms, e mail, and endpoints, Zscaler provides best-in-class instruments to safe information on this quickly evolving panorama, offering visibility, management, and Zero Belief safety for enterprise functions worldwide. This permits enterprises to:
- Discover delicate information throughout endpoints, inline, and cloud with AI-powered auto information discovery and classification.
- Shield information in movement with full TLS/SSL inspection and inline DLP for internet, e mail, BYOD, and GenAI apps.
- Safe information at relaxation in clouds and on endpoints with unified coverage, sharing controls, and machine posture.
- Simplify operations with unified end-to-end incident response utilizing a single, built-in console with Workflow Automation.
Defending enterprise AI apps from information loss
Zscaler additionally delivers a full suite of best-in-class merchandise to safe generative AI instruments like ChatGPT and Microsoft Copilot.
- AI app visibility: As staff quickly undertake AI instruments like ChatGPT and Microsoft Copilot, Zscaler ensures enterprises by no means lose visibility over delicate inputs or outputs.
- Sensible enter immediate blocking: Zscaler makes use of AI/ML-driven URL filtering and coverage enforcement to categorize AI app exercise and mechanically block unsafe or unapproved enter prompts.
- Deep visibility into AI workflows: Progressive categorization of person prompts lets safety groups monitor, analyze, and make educated choices about AI utility safety. For example, Zscaler insurance policies can:
- Monitor for delicate person information (e.g., social safety numbers) in actual time.
- Block prompts associated to mental property leakage.
- Safe collaboration through isolation: Forestall unintentional information transfers in AI functions, with out stifling productiveness:
- Browser isolation for AI instruments: Zscaler’s Browser Isolation know-how permits staff to work together with AI instruments securely by rendering functions in an remoted digital browser.
- Clipboard utilization, file uploads, and downloads might be restricted whereas nonetheless enabling prompts.
- Forestall unintentional information exfiltration when staff work together with generative AI apps, reminiscent of ChatGPT or OpenAI-powered interfaces.
- Browser isolation for AI instruments: Zscaler’s Browser Isolation know-how permits staff to work together with AI instruments securely by rendering functions in an remoted digital browser.
- Secure pixel rendering: By rendering functions as “pixels,” Zscaler ensures delicate data by no means bodily leaves the group’s management, even throughout distant use.
- Securing Microsoft Copilot: With Microsoft Copilot set to revolutionize enterprise productiveness, Zscaler eliminates dangers tied to delicate information misuse, misconfigurations, and third-party entry.
- Inline information leak prevention for prompts: Zscaler scans OneDrive information and Copilot features in actual time, mapping information connections to make sure safety requirements. Forestall extra permissions and proactively block delicate information from publicity.
- Repair misconfigurations in SaaS settings: Zscaler repeatedly screens configurations to resolve oversharing dangers.
- Finish Consumer Behavioral Analytics (EUBA): Utilizing AI-driven behavioral analytics, Zscaler identifies anomalies not solely from Copilot customers but additionally from any linked third-party SaaS integrations.
There has by no means been a extra crucial time to rethink your enterprise’s strategy to information safety. The 2025 ThreatLabz Knowledge@Danger Report provides a complete have a look at the place dangers lie, what drives them, and the way organizations can reply successfully to safe their delicate information in at this time’s quickly evolving, AI-driven ecosystem.
For a full record of finest practices, obtain the 2025 Knowledge@Danger Report.
