Artificial intelligence has become a decisive element that benefits both attackers and defenders. In this scenario, José María Alonso, Snowflake’s country manager for Spain and Portugal, highlights three major trends that will shape the evolution of this particular technological race through 2026.
Cyber Agents Will Become Weapons in the Next Wave of Cybercrime
The cybersecurity arms race has always been characterized by the constant tug-of-war between attackers and defenders, but the emergence of AI agents capable of researching, designing, and executing attacks will dangerously tip the balance. By 2026, agentic cybercrime will become a frontline issue, with defenders facing a new class of adversary. One of the biggest risks with AI agents will be prompt injection — adversaries deceiving systems to bypass security barriers — and the hallucinations they produce, leading to false or misleading results. We can expect to see agents that analyze code, identify vulnerabilities, and build customized exploit kits to exfiltrate data and deploy ransomware. Additionally, there will be cases where AI creates fake sales documents or security claims, putting companies at risk of legal penalties. But this is just the beginning. The real turning point will occur when agents stop merely mimicking attackers and begin devising entirely new strategies, at which point defenders will face a completely new level of challenges.
Cybercriminals Will Leverage Dark AI to Escalate Attacks
While current foundational models are built with security barriers in place, malicious actors are already deploying uncensored versions like FraudGPT and WormGPT to generate phishing campaigns, malicious code, and social engineering attacks. In many ways, this represents the dark side of open-source code, with malicious actors taking open-source models like GPT-J-6B and deploying them without the ethical safeguards of commercial systems. AI-enhanced tools will quickly become part of the supply chain that drives cybercrime-as-a-service. This underground economy will no longer depend on individual attackers but on global companies packaging and selling complete cybercrime infrastructures with subscription levels, customer support, and regular updates. As these offerings mature, even the most advanced and costly AI models will inevitably be weaponized.
AI Tools Will Address the Talent Shortage in Cybersecurity
Despite all the risks introduced by AI, it also holds a real promise for defenders. Generative and agentic AI will start providing security operations centers with the scale they have lacked. The most persistent challenge for CISOs has been the shortage of qualified analysts. Security talent is hard to find and retain. Rather than replacing human expertise, advanced AI will fill the security gaps that have gone unstaffed for years, enhancing analysts’ capabilities and leveling the playing field. Over the next three years, AI agents will provide the necessary force multiplier to finally shift the balance back in favor of defenders.