Departing employees still email files to personal accounts and copy them to USB drives. Now they also paste sensitive content into ChatGPT, Claude, or Gemini to summarize a codebase, draft a portfolio piece, or package a project before their last day. Neither channel has replaced the other. The attack surface has simply gotten wider, and most security teams are still staffed and tooled for the channels they already know how to watch.
That expansion is what makes departing employee risk harder to manage today. Access is already intact, motivation shifts before an exit, and AI tools give departing employees one more fast, low-friction way to reformat and carry out data alongside the file transfers and email attachments security teams have monitored for years.
What Is Departing Employee Data Exfiltration?
Departing employee data exfiltration refers to the unauthorized removal of sensitive company data by an employee preparing to leave or already resigned. This insider threat action, which can involve source code, customer records, or strategic documents, occurs deliberating and is often rationalized as harmless, despite sensitive data leaving the organization. Increasingly, this exfiltration may also involve pasting that sensitive data into an AI tool to summarize or repurpose it, which moves the data outside company control just as simply as an email attachment does. Departing employees are consistently one of the most common types of insider threats.
The 30-Day Window Around Resignation
Departure risk is not evenly distributed across an employee's tenure. Cyberhaven data shows that office-based employees are 77 percent more likely to exfiltrate sensitive data than remote employees, and that risk climbs sharply in the 30 days before and after a resignation is submitted.
That window is exactly when AI-assisted exfiltration is hardest to catch, since the employee's day-to-day AI usage does not visibly change. The same tools they used all year for drafting and summarizing are now being pointed at higher-value data, and the volume and pattern of that activity, not any single event, is the signal worth watching.
Why Departing Employees Are Turning to AI Tools to Take Data With Them
An employee who has accepted a new role does not need to zip up a folder and email all the organization’s data home. They can ask an AI assistant to summarize a technical design document, rewrite a client presentation in general terms, or turn a year of project notes into a portfolio sample, then carry the output forward into their next job. The company's data went in. Something derived from it comes out, often on a personal device or a personal AI account with no enterprise data retention controls.
This is not always malicious. Many departing employees genuinely believe they are entitled to work they personally produced, or they simply want a reference for their next role. Cyberhaven Labs research shows that 39.7 percent of all AI interactions involve sensitive data, and the same behavior that shows up across the workforce day to day becomes concentrated risk in the weeks around a resignation, when the data being summarized is often proprietary rather than incidental.
How Legacy DLP Misses AI-Driven Exfiltration
Traditional data loss prevention (DLP) tools were built to catch data leaving through email, file transfers, and removable media. Pasting a paragraph into a browser-based AI tool does not trip any of those detections. There is no attachment, no download, and often no file movement at all, just clipboard activity inside a browser tab.
Legacy DLP tools that rely on content inspection also cannot connect that paste event back to the original file it came from. A confidential product roadmap pasted into an AI prompt, or pulled into a folder by an AI agent looks like unstructured text or a normal workflow, not a policy violation, unless the tool understands where that text originated and how sensitive it is. Without that context, the paste event is either ignored or buried among thousands of other browser alerts that turn out to be nothing.
How Data Lineage Detects AI-Driven Insider Risk
Consider a departing employee who spends their final two weeks pasting portions of internal documentation into an AI tool to draft a writing sample for their next interview. Most detection tools would only see the activity: a browser session, a paste event, nothing unusual on its own. Data Lineage looks at the data instead, tracking where that documentation originated, how it has moved and transformed, and who has touched it along the way, regardless of which application or AI tool is carrying it at the moment.
That difference matters because the same paste event looks identical whether the employee pasted public onboarding material or a confidential product roadmap. Data Lineage closes that gap by preserving the connection to the original source even after the text is copied, reformatted, or renamed, so the writing sample the employee generates can still be traced back to the confidential document it came from, something content-inspection tools cannot do.
That data-level context is what turns a routine-looking paste event into an actionable one. Instead of an alert that only says text was pasted into an AI tool, a security team gets the full record: what the document was, how sensitive it is, and why a departing employee touching it in the final weeks before their last day is worth a closer look.
How Cyberhaven Addresses Departing Employee Risk
Cyberhaven's insider risk management (IRM) capability combines behavioral signals, such as job search activity and unusual file access, with Data Lineage's understanding of the data itself, giving security teams visibility into departing employee activity across endpoints, SaaS applications, cloud platforms, and AI tools. Rather than only alerting when a departing employee mishandles data, Cyberhaven can intervene to block the transfer across channels, including cloud storage, email, and browser-based AI tools.
When an employee copies a section of a confidential document and pastes it into an AI assistant, Cyberhaven traces the action back to the source file, flags its sensitivity, and surfaces the event with context instead of a bare alert. If an investigation opens, security teams get a pre-assembled chain of custody covering both traditional channels and AI tool activity, including what data an employee accessed, downloaded, or pasted throughout their notice period, rather than spending days reconstructing it from separate logs.
Learn more about how to stop departing employees from exfiltrating data.
.avif)
.avif)
