Insider threats have traditionally conjured images of bad-intentioned employees stealing sensitive information or abusing legitimate access, but as organizations mark Insider Threat Awareness Month this September, security leaders are warning that the definition of an “insider” is rapidly expanding.
Generative AI, deepfakes and autonomous agents are creating new ways for trusted access to be exploited, sometimes without a malicious employee being involved at all. Attackers can impersonate executives or candidates, employees can inadvertently expose sensitive data through AI tools, and autonomous agents can take damaging actions at machine speed.
One emerging concern starts before an employee even joins an organization. Advances in synthetic identities and deepfake technology are making it easier for attackers to pose as legitimate candidates, potentially turning the hiring process itself into an entry point.
“GenAI, deepfakes and synthetic identities have fundamentally changed the hiring threat model. Bad actors can now pass an interview, get hired and receive legitimate credentials, without having to break in,” said Bojan Simic, CEO and co-founder of HYPR. “The security failure happens when we authenticate a credential without first establishing that it belongs to the right human.”
Simic said organizations should extend zero-trust principles to “Day Zero,” tying identity verification directly to credential provisioning rather than relying solely on interviews, background checks or human intuition.
Deepfakes are also creating risks once employees are inside the organization. Attackers can use publicly available information about executives and company processes to create convincing impersonations designed to manipulate legitimate employees.
“Executive impersonation is a growing insider threat method, as attackers can now combine publicly available video with knowledge of an organization’s leadership, processes and culture to create highly convincing requests,” said Amit Shuster, VP of Product and Engineering at Vetric. “A familiar face and voice can make an urgent instruction from a CEO to transfer funds, disclose sensitive information or bypass a control feel legitimate.”
Shuster said organizations need verification processes that do not depend solely on whether a video appears authentic, while investigators need greater visibility into how impersonation campaigns are created and distributed.
But some of the most common AI-related insider risks may be considerably less sophisticated. Employees increasingly use public or unauthorized GenAI tools to work faster, potentially exposing corporate information in the process.
“When we talk about insider threats, most organizations still picture a disgruntled employee stealing files on their way out the door. The bigger risk today is far less dramatic and far more common: well-meaning employees pasting sensitive data, source code, or customer information into GenAI tools because it's faster than the approved workflow,” said Arti Raman, CEO of Portal26.
Rather than attempting to prohibit AI entirely, Raman argued that organizations need visibility into how employees are actually using the technology so security teams can identify risky behavior and steer users toward safer alternatives.
Detecting these different forms of insider activity also requires organizations to connect signals that may be scattered across numerous systems.
“Insider threat activity often originates from people and systems that are already trusted. The warning signs rarely live in a single alert or data source,” said Mike Wade, VP of Customer Success at Gravwell. “They emerge when security teams can connect activity across identity, network, endpoint, cloud and other telemetry over time.”
Wade said retaining diverse, full-fidelity telemetry can become particularly important during investigations because security teams cannot always predict which information will ultimately prove significant.
At the same time, autonomous AI is introducing an entirely different category of insider risk: a trusted non-human identity capable of interacting with corporate systems and data.
“Organizations are giving AI agents broad access to corporate systems and data, and unlike human insiders, they move at machine speed,” said Arvind Parthasarathi, CEO and founder of CYGNVS. “They don't even need malicious intent to cause serious damage. An agent pursuing the wrong objective — or the right objective without proper guardrails — can create a security, legal or regulatory crisis before the human response team even understands what happened.”
Parthasarathi said organizations should establish response procedures before agents cause an incident, including determining who has authority to stop an agent, how it should be contained, what evidence needs to be preserved and which legal or regulatory obligations could be triggered.
Taken together, the risks point to an insider threat landscape increasingly defined by trusted access rather than simply trusted people. Whether that access belongs to an employee, an attacker posing as one or an autonomous AI agent, organizations will need to rethink how identities are verified, behavior is monitored and incidents are investigated as AI becomes more deeply embedded in everyday operations.
