AI is reshaping cybersecurity — arming attackers as fast as defenders
Generative AI has become a powerful tool for cybercriminals, making sophisticated attacks accessible even to inexperienced actors. Security professionals are deploying AI defenses in response, but attackers still hold a significant edge.
Artificial intelligence language models have fundamentally altered the cybersecurity landscape, allowing even inexperienced criminals to carry out attacks that once required years of expertise. Experts say AI has effectively "democratized" cybercrime, since these models can synthesize knowledge across disciplines including coding, networking and social engineering.
Speed and scale
AI lets attackers accomplish in hours what once took weeks. In one known case from February, attackers who breached Mexican government databases used multiple AI models simultaneously, passing outputs between them to work around restrictions built into any single system.
Phishing and social engineering
Phishing has been particularly transformed by AI: messages are now grammatically flawless and highly personalized, as chatbots can independently research a target online. According to security vendor Brightside, roughly 82 percent of phishing emails now involve AI at some stage, and AI-assisted messages have pushed click-through rates in simulated tests from 12 percent to 52 percent. Voice cloning and deepfake attacks have also surged sharply.
New attack surfaces
As organizations connect more systems to AI, they open new vulnerabilities such as prompt injection attacks, where malicious instructions hidden on a webpage are unknowingly executed by an AI assistant. In one case, a support chatbot Meta introduced in December 2025 made it trivially easy for attackers to hijack Instagram accounts. Experts also warn of a theoretical threat known as data poisoning, in which malicious content is deliberately seeded into sources used to train AI models.
Defenders fight back
The cybersecurity industry is deploying AI tools of its own to strengthen defenses, but the fight remains uneven — defenders must protect every possible weakness while attackers need only find one. Surveyed security professionals ranked defending against fraud and deepfakes, along with application-level attacks, as top priorities, while cloud environments were cited as the hardest surfaces to secure.
