CISA updated its Insider Threat Mitigation Guide on Sept. 9, expanding a framework first published in 2020 to account for how much the workplace has changed. The update gives more attention to remote work and to newer risks such as the use of AI to deceive employees.
It also expands guidance around who can enter sensitive spaces and what happens when an employee leaves under difficult circumstances. The changes build on CISA’s existing approach to insider threats rather than redefining them.
“Insider threats continue to evolve as technology becomes more advanced. We urge organizations to establish a mitigation program that protects key assets, prevent violence, reduce losses, safeguard sensitive data, and save lives,” said Acting Executive Assistant Director for Infrastructure Security Scott Breor in CISA’s announcement of the update.
The underlying problem the guide addresses is still trusted access. An employee, contractor, or other authorized user can create risk intentionally, by mistake, or after an outside attacker takes control of an account. As work has become more distributed, determining when legitimate access has become dangerous has grown more complicated.
A 2025 report from Fortinet and Cybersecurity Insiders found that 77% of surveyed organizations had experienced at least one insider-related data-loss incident during the previous 18 months. Twenty-one percent had detected more than 20.
Trusted Access No Longer Stops at the Office Door
CISA’s update gives more attention to the way hybrid and remote work have changed insider risk. Trusted access increasingly reaches beyond company facilities, making it harder for organizations to see the full context around legitimate activity.
A valid login or routine file transfer may be exactly what it appears to be, or it may signal that an account has been misused.
“We see these breaches consistently emerging as a significant, common challenge across all industries,” said Rex Booth, chief information security officer at SailPoint. “This frequency makes sense because modern enterprises rely heavily on granting broad access to empower dynamic human workflows.”
AI Creates a New Deception Problem—and a Potential Blind Spot
CISA’s update adds the use of AI to manipulate or deceive employees. A convincing synthetic voice, message, or video can persuade someone to disclose information, approve access, share credentials, or take another action on an attacker’s behalf. In those cases, an employee may become an unwitting participant in the attack.
AI can also amplify familiar insider risks. An attacker who compromises an employee’s account can operate through legitimate access, while workers themselves may expose company data through unsanctioned AI tools.
The guide largely focuses on attackers using AI to manipulate or deceive employees. It says less about AI systems that operate inside an organization with delegated access and permissions. That raises a different concern as companies give autonomous systems greater access to internal resources.
“Organizations are rapidly deploying autonomous systems with broad permissions to access data, APIs, and automate business workflows,” said Morey Haber, Chief Security Advisor at BeyondTrust. “If improperly governed, lacking secure by design principles, these identities can unintentionally expose sensitive information or execute actions beyond their intended scope.”
Managing Risk When Access Is Changing or Ending
CISA also adds guidance for what it calls adverse employee separations, or departures where the circumstances raise the risk of retaliation or misuse.
The emphasis is on controlling access when someone is leaving. In some cases, that may mean cutting off access immediately. In others, an employee may continue working for a period under tighter restrictions while a departure or investigation is underway.
The challenge for defenders is that access does not exist in only one place. Disabling a user account does not solve the problem if building access remains active, and recovering a laptop does not end access to cloud services. CISA’s guidance points toward treating departure as one coordinated process rather than a series of disconnected tasks.
The same logic applies to visitors. Temporary access should reflect what a person needs while on-site and end when that need is over.
More broadly, insider-risk controls have to account for changes in access over time. Waiting until after an incident can leave an organization trying to reconstruct what should have been restricted earlier.
From Warning Signs to a Working Insider-Risk Program
CISA organizes insider-risk programs into four phases: define the threats an organization faces, detect and identify possible threats, assess their level of risk, and manage them.
No single department has the full picture. Security teams may see unusual account activity without knowing whether it fits the person’s role or circumstances. That context matters because unusual activity is not necessarily malicious.
Access controls and monitoring can reduce opportunities for misuse, but they cannot eliminate human error or prevent every successful social-engineering attempt. Employees also need a clear way to report suspicious activity or their own mistakes before a small problem becomes a larger one.
“The goal is not blanket employee surveillance,” said Aviv Nahum, Co-founder and CEO at Above Security. “It is to build enough context to intervene proportionately—sometimes that means investigation, sometimes coaching, sometimes restricting access—before a concern becomes a breach.”
A Starting Point, Not a Complete Insider-Threat Model
For organizations without a mature insider-risk program, CISA’s guide offers a framework and supporting resources for putting one into practice. More advanced programs may need additional controls for systems that do not fit neatly into a human-centered definition of an insider.
Insider risk is an ongoing access-management problem. Organizations need to keep evaluating who—or what—has trusted access and whether that access is still necessary.
