Overreliance on artificial intelligence can turn users into potential targets for cybercriminals. Individuals and companies, in the pursuit of increased productivity and speed, delegate a significant portion of their activities to AI agents without carrying out necessary checks.
Blind trust in artificial intelligence can significantly increase cybersecurity risks. Today, the most important security weaknesses are no longer merely in software or hardware, but in how people interact, especially with AI agents.
AI capabilities are developing rapidly, and these systems can now plan, use APIs and control tools, and perform complex tasks automatically and without direct human intervention.
However, these advances also carry the risk of complete reliance on convenience and productivity.
According to a researcher at Kaspersky, ideally every stage of AI program development should be reviewed, but real-world oversight of these processes is often neglected due to a focus on speed and meeting deadlines. When trust in AI becomes blind trust, a significant cybersecurity risk arises.
Cybercriminals no longer necessarily need to identify technical complexities to carry out attacks; they can exploit users' overreliance on AI.
This threat does not only target users and also extends to the core infrastructures of technology companies that rely on open-source software. Developers and AI agents depend on these software packages to perform their tasks, and that dependency can become a point for executing attacks.
According to Kaspersky’s report, 31 percent of companies have experienced attacks that targeted their supply chain. One example of such an attack was against the JavaScript library Axios, during which attackers were able to take control and deploy a malicious version of the software.
In addition, Kaspersky Container Security has identified thousands of potential misconfigurations in CI/CD processes on the GitHub Actions platform.
The persistence of these threats indicates that continuous verification, review, and monitoring of AI programs and agents is of great importance for maintaining their safety and security.