Cybercriminals are increasingly leveraging artificial intelligence (AI) agents and chatbots to autonomously orchestrate and execute cyber-attacks, according to a comprehensive analysis of hacker behavior conducted by cybersecurity researchers at ESET. This alarming trend signals a significant shift in the tactics employed by cybercriminals, as they harness advanced technologies to enhance their malicious activities.
ESET’s recent threat report, which focuses on findings from the first half of 2026, mentions that the researchers scrutinized approximately 900,000 AI skills—small functional components utilized by AI agents. These skills, found in public repositories, revealed a troubling surge in both suspicious and outright malicious instances. The report, published on July 8, underscores that the proliferation of these AI toolsets has broadened the attack landscape for cybercriminals, subsequently increasing the vulnerability of organizations to cyber threats.
AI agents possess the capability to execute a multitude of tasks autonomously. They can browse the web, interact with third-party services, write files, execute commands, and perform various actions on behalf of users. While legitimate users frequently utilize AI agents to enhance their productivity, this growing trend has also attracted cybercriminals and other malicious actors eager to exploit these technologies to their advantage.
The analysis conducted by ESET highlighted a stark increase in the number of suspicious and malicious AI tools in recent months. The number of AI agent skills flagged as suspicious skyrocketed from around 10,000 to over 25,000 during the reporting period. Simultaneously, the count of skills deemed outright malicious grew embarrassingly from approximately 600 to over 3,000. This escalated availability of nefarious AI tools illustrates an evolving threat landscape brought about by advancements in technology.
### Exploiting AI Tools for Malicious Intent
The investigation revealed that these AI agent toolsets are either misused or specifically designed to perform actions on behalf of malicious attackers. Some tools are intentionally placed in open-source repositories, with the hope that unsuspecting users will inadvertently download these malicious toolsets. Others are sold directly to criminal enterprises as outright malicious products.
The malicious capabilities enabled by these AI tools are numerous and deeply concerning. They allow attackers to exfiltrate sensitive data, download and execute malware, override user instructions, and even subtly alter the behavior of the AI agents themselves. For instance, a set of tools originally marketed for legitimate “red teaming” activities was found to have functionalities that far exceeded its advertised capabilities, including the ability to exfiltrate user credentials and maintain persistent access to targeted systems. Some malicious tools were even capable of deploying remote access tools like Mimikatz, which is notoriously linked to ransomware attacks.
In addition to fully malicious tools, there exists a substantial number of applications that, while not outright harmful, possess structural vulnerabilities that cybercriminals can exploit for nefarious purposes. The trend echoes previous tactics employed by cybercriminals, who have previously developed harmful browser extensions and mobile applications for malicious ends. However, ESET cautioned that the incorporation of AI into these tools could exacerbate the threats posed to potential victims.
The ESET report emphasizes a critical point: “When it comes to handling sensitive data, making purchases, running API calls, or instruction chains, the higher level of autonomy of AI agents increases the risk and scope of such attacks.” This underscores the potential for significant data breaches due to the unregulated utilization of AI in cybercriminal operations.
Given this escalating landscape of cyber threats, organizations must actively re-evaluate their cybersecurity strategies. It becomes imperative for businesses to establish safeguards against the use of potentially harmful tools and to raise awareness among employees about the dangers of downloading free tools from unverified sources.
Jake Moore, a global cybersecurity advisor at ESET, highlighted the urgency of cautious practices, stating, “Although AI uses impressive speed and autonomy, we can still do our best in protecting data with familiar defences, such as recognizing tools seeking excessive access to files or credentials for trivial tasks, and being wary of anything pushed through hype rather than official channels.”
In a world where free AI tools promise extraordinary capabilities in exchange for sweeping access, organizations and users alike must remain vigilant and critical of the tools they choose to employ in their daily operations. This vigilance is crucial in safeguarding sensitive information and preventing potential breaches that could have devastating consequences. The necessity for a proactive approach in understanding and mitigating the risks associated with AI technologies has never been greater.