On June 1, 2026, La Vanguardia published an interview with Richard Marko, CEO of ESET, conducted during the ESET World 2026 conference, ESET’s annual cybersecurity event held in Berlin on May 19 and 20. In it, Marko summed up the current state of cybersecurity in one sentence: «With AI capabilities, it is easier to use them to attack than to defend». This warning, coming from the head of one of the world’s oldest cybersecurity companies, deserves careful analysis by any IT or business leader in Spain.
A new era: the CEO of ESET calls it this and has reasons
Richard Marko has been in the cybersecurity sector for decades. He has seen floppy disk viruses, email worms, ransomware, and supply chain attacks pass by. And now he states unequivocally that we are facing a paradigm shift.
The reason is specific: with current artificial intelligence tools, an attacker can design malware faster and more efficiently than ever before. Not only that. According to Marko, there are already autonomous systems capable of setting a target and trying different attack strategies until they find one that works, adapting to the obstacles they encounter along the way.
This represents a qualitative leap compared to any previous threat:
- Past: an attacker needed advanced technical knowledge to create functional malware. The process took days or weeks.
- Present: with generative AI tools, the same process can be carried out in hours, with less technical expertise, and produce virtually unlimited variants that evade known signatures.
The asymmetry is brutal: the attacker only needs one attack to succeed. The defender needs to stop all.
Prompt injection: when malware is just text
One of the most unsettling concepts Marko addressed in Berlin is that of prompt injection: a new category of attack that exploits language models and AI agents that more and more companies are integrating into their workflows.
Unlike traditional malware — executable code, scripts, macros — a prompt injection attack is, literally, text. Instructions written in natural language that, when read by an AI system connected to business tools (email, calendar, CRM, ERP), can manipulate it to execute unauthorized actions: exfiltrating data, forwarding emails, modifying records, or opening backdoors.
What makes it especially dangerous:
- It does not look like malware. Any analyst can read it and may not detect the threat.
- It can be written in any language, including languages uncommon to Western security teams.
- It does not trigger traditional defenses based on executable file signatures.
- It spreads through documents, emails, web pages, or any text source processed by the AI agent.
For Spanish companies adopting AI assistants integrated with Microsoft 365, cloud CRMs, or automations with AI agents, this threat is real and growing.
What does this mean for a Spanish SME?
Marko’s warning is not directed only at large corporations or critical infrastructure. SMEs are an especially attractive target in this new scenario for several reasons:
- Lower investment in cybersecurity: many companies with 10-100 employees still use consumer antivirus or solutions without centralized management.
- Lower response capacity: without a dedicated security team, the time to detect and respond to an incident is much longer.
- Higher profitability for the attacker per effort: with AI tools that automate attacks, attacking 1,000 SMEs costs almost the same as attacking one. Volume compensates.
- Dependence on cloud tools: the massive adoption of Microsoft 365, Google Workspace, and various SaaS solutions greatly expands the attack surface.
The CCN-CERT Cyber Threat Report for 2025 already pointed out that 60% of serious cyber incidents in Spain affected organizations with fewer than 250 employees. AI has democratized access to sophisticated attack tools: you no longer need to be a nation-state to launch advanced attacks.
The response: AI against AI, but trained for cybersecurity
Marko’s conclusion in Berlin is not pessimistic, although it is realistic: «We always find a way to deal with a new situation». And the way to deal with this is clear: using AI in cyberdefense, but in a modern way.
What does “modern way” mean? It is not enough to add the “AI” label to an existing antivirus. Marko speaks of AI models trained specifically with cybersecurity data: millions of real malware samples, logs of compromised system behavior, attack patterns documented over decades.
This is precisely the basis of ESET’s architecture: since the nineties, ESET has built one of the largest threat analysis laboratories in the world (ESET Research), whose data feeds the machine learning models behind all ESET PROTECT products. It is not generic AI. It is AI trained with three decades of real threats.
How does this translate into ESET solutions for businesses?
- ESET PROTECT Entry: multi-layer protection with machine learning and real-time behavior analysis. The engine detects previously unseen malware variants based on behavior, not just known signatures.
- ESET PROTECT Advanced: adds ESET LiveGuard Advanced, the cloud sandbox that runs suspicious files in an isolated environment before allowing them to reach the endpoint. Especially effective against polymorphic malware generated by AI.
- ESET PROTECT Complete: incorporates native protection for Microsoft 365 (Exchange Online, SharePoint, OneDrive, Teams), the most exploited entry vector in 2025-2026 for AI-driven phishing and prompt injection attacks.
Five measures your company can take right now
Beyond the security solution, the new scenario requires a broader approach:
- Update your endpoint protection solution to one that includes behavior analysis and sandboxing. Antivirus based solely on signatures are insufficient against AI-generated malware.
- Protect cloud email. If you use Microsoft 365, you need a specific security layer for Exchange Online. Microsoft’s native spam filter is not enough.
- Train your employees on AI phishing. AI-generated phishing emails are grammatically correct, personalized with real information from LinkedIn or social media, and practically indistinguishable from legitimate communications.
- Establish policies for the use of AI agents. If your company uses tools with integrated AI (Copilot, ChatGPT Enterprise, automations), define what data they can process and what actions they can execute autonomously.
- Enable multi-factor authentication (MFA) on all services. 80% of unauthorized access to corporate accounts occurs due to compromised credentials. MFA blocks them even when the password is known.
Frequently asked questions about AI and enterprise cybersecurity
Why does AI favor attackers more than defenders?
According to Richard Marko, CEO of ESET, with AI it is possible to design malware faster and more efficiently, and create autonomous systems capable of trying different attack strategies and adapting. The attacker only needs one attack to work; the defender must anticipate all possible vectors simultaneously.
What is prompt injection and why is it dangerous for businesses?
Prompt injection is a new form of attack that exploits AI agents and models integrated into business tools. They are malicious instructions written in plain text that can manipulate an AI system to execute unauthorized actions: exfiltrating data, forwarding information, or opening access points. It is difficult to detect because it looks like normal text and can be written in any language.
How can a company protect itself against AI cyberattacks?
By using AI in cyberdefense in a modern way: with models trained specifically with cybersecurity data. Solutions like ESET PROTECT incorporate machine learning and behavior analysis trained on decades of real threats, detecting new malware that traditional signatures do not recognize. The cloud sandbox (ESET LiveGuard Advanced) is especially relevant against new variants generated by AI.
What is ESET World 2026?
ESET World 2026 is the annual cybersecurity conference of ESET, held in Berlin on May 19 and 20, 2026. At this event, CEO Richard Marko presented the company’s vision on the new era of AI-based cybersecurity, warning about the asymmetry between attack and defense with artificial intelligence tools.
Which ESET solution is most suitable against AI threats?
For most Spanish SMEs, ESET PROTECT Complete offers the optimal balance: cloud sandbox, Microsoft 365 protection, disk encryption, and patch management. For organizations with greater exposure, ESET PROTECT Elite adds full XDR capabilities with extended detection and response.
Do you want to evaluate if your company is prepared for the current threat scenario with AI? Contact us — as authorized ESET distributors in Spain, we offer a free diagnosis of your current security environment and recommend the most suitable solution without commitment.
Source: Richard Marko: «It is easier to use AI capability to attack than to defend» — La Vanguardia, 06/01/2026. Interview by Francesc Bracero, special correspondent at the ESET World 2026 in Berlin (May 19-20, 2026).
