According to recent reports, 2026 is widely regarded as one of the most perilous years for internet users, as cybercrime continues to evolve through the use of artificial intelligence. Researchers note that AI tools are no longer limited to routine automation but are now used to enhance cyberattack techniques such as phishing and social engineering. By combining large language models with deepfake technologies, attackers can convincingly impersonate trusted individuals, making scams more difficult to detect and resist. One particularly concerning trend is the rise of highly personalized scams, including so-called "pig butchering" schemes. These schemes involve criminals building emotional relationships with targets over extended periods before introducing fraudulent investment opportunities. With AI, such operations can be carried out at scale while maintaining a high degree of personalization, blurring the line between automated and human interaction. Cybersecurity experts also warn of a potentially more disruptive phase in AI-driven risks. A leaked blog post from Anthropic described its upcoming model, Mythos, as a possible "watershed moment" with serious security implications. The model is believed to have the ability to identify and exploit system vulnerabilities at unprecedented speed, raising concerns about large-scale cyberattacks if such technologies are realized. Advances in autonomous malware add further risk. Some AI-driven systems can modify their own code during execution, allowing them to evade detection and adapt to defensive measures. Experts stress that these developments not only increase the scale of cybercrime but also make it more sophisticated and unpredictable. Although AI can strengthen cybersecurity defenses, experts highlight a key imbalance: attackers need only find one vulnerability, while defenders must secure entire systems. This suggests that digital security will become increasingly complex and difficult to manage.