Attackers are increasingly using trusted AI applications, legitimate IT tools and stolen credentials to gain access to enterprise systems, according to a report by Elastic Security Threat Command, which analysed confirmed intrusions and research conducted between April and September 2026.
Researchers observed attackers executing commands through AI applications including Claude Code on Windows and ChatGPT desktop on macOS, exploiting the system-level access and credentials available to these tools. In one incident, a command executed on a macOS device accessed an internal infrastructure document and retrieved saved passwords from the user’s keychain.
The report also found that attackers used stolen credentials an average of eight minutes after malware infection across 27 confirmed cases. ClickFix attacks, which trick users into executing malicious commands, accounted for 12 cases, or 44% of the incidents. In one healthcare case, malware was delivered approximately 16 seconds after a user fell for a malicious prompt.
“Attackers hiding inside trusted software isn’t new. Defenders have been fighting living-off-the-land techniques for decades, and MITRE ATT&CK has an entire tactic built around it,” said Mike Nichols, general manager, Security, Elastic.
“What’s new is that enterprises are increasingly deploying powerful trusted tools that most security teams struggle to fully monitor, control, or even explain: AI agents. An agent that holds credentials and can reach the internet is a new attack surface, whether it’s misconfigured or abused. When an attacker is behind it, its activity looks exactly like the employee who installed it. So the question for defenders isn’t whether a tool is approved; it’s what that tool is actually doing.”
Researchers also reconstructed the July 2026 Hugging Face breach, identifying signs of AI-agent activity such as repeated commands executed within seconds and automated actions continuing through errors. Separately, the report confirmed 23 cases involving attackers installing legitimate remote IT software to maintain access to compromised devices, most commonly ScreenConnect.
Elastic recommends monitoring trusted tools based on their behaviour rather than approval status, tracking what AI applications execute and where remote management tools connect, and correlating endpoint, cloud and identity signals to identify suspicious activity.