Artificial intelligence and new cyber threats: how can this system be protected from threats?

DUSHANBE, 13.09.2026 /AMIT "Khovar"/. As artificial intelligence develops, cyber-defence capabilities expand, but a new layer of threats also emerges. The report says this technology does not belong only to defenders: attackers also use it to increase the speed, scale and effectiveness of phishing, social engineering, malware development, cyber espionage and the automation of some stages of an attack.

The report states that in today’s cyber environment, AI is both a defence tool and an attack tool. The key issue is not only how powerful AI is, but how safely it is used and how the AI system itself is protected from threats. The text also distinguishes between "AI for Cybersecurity" and "Security of AI".

It further highlights risks in LLM and GenAI systems, including prompt injection, sensitive information disclosure, supply chain attacks, data or model poisoning, excessive agency and leakage of system prompts. The conclusion is that any organisation introducing AI into critical processes must also put in place technical protection and risk management for the AI system itself.