AI: from support to autonomous execution of cyberattacks

iGuRu.gr > AI: from support to autonomous execution of cyberattacks

Η Check Point Software Technologies Ltd. cybersecurity solutions provider published its annual report AI Security Report 2026 Check Point Research, revealing a defining shift in the cyberthreat landscape: over the past twelve months, artificial intelligence has gone from supporting attackers to carrying out attacks itself. Where the AI used to help cybercriminals prepare, it now actively participates in executing attacks with minimal human intervention, significantly reducing the reaction time of defenders and creating new risk surfaces for businesses. AI Security Report 2026

The report draws on real-world incidents, telemetry, and original case studies from the past year, demonstrating how AI is now directly involved in every stage of a cyberattack chain.

See more articles from iGuRu.gr when you search for news on Google.

iGuRu.gr preferred source

Key findings of the report

AI is now executing attacks , not just supporting them. Researchers have documented attacks in which AI autonomously managed vulnerability exploitation processes, executing thousands of commands across dozens of sessions with minimal human guidance.
The “window of vulnerability” has shrunk from days to hours. AI can now turn a newly disclosed vulnerability into a functional exploit within hours, drastically reducing the time it takes for authorities to restore critical systems.
Detections of malicious prompt injections increased approximately fivefold between March and May 2026, demonstrating that indirect manipulation of AI models is now a real operational threat rather than a theoretical risk.
Identity can no longer be considered a sufficient security mechanism on its own. Voice, face, documents and videos can now be convincingly faked by AI , while even expert evaluators correctly identified only about 41% of faces created by AI.
High-risk interactions with AI applications doubled in a year , from one in 50 to one in 25. At the same time, the average business now uses ten different AI applications each month, many of which are unapproved.
The greatest exposure of corporate data comes from legitimate use of AI tools , not attacks , as employees often share more information than they realize in order to get useful answers.

Lotem Finkelstein, Vice President at Check Point Research , said:

“A year ago, we described AI as a force multiplier for attackers. This year, we see something even more significant: AI is now actively integrated into the attack chain, capable of executing operations that previously required a specialized team. The barrier of expertise that separated skilled attackers from the rest is rapidly decreasing, and defenders can no longer assume that humans set the pace of the attack. The organizations that will stay ahead will be those that establish rules for the use of AI , protect the AI systems they depend on, and operate at the speed of a machine, not a human.”

What can organizations do?

The report proposes three main axes of defense against modern AI threats :

Security for AI : Protecting the very AI systems that organizations use.
Security by AI : Leveraging AI to prevent and respond to threats at machine speed.
Security with AI : Governance of AI use by human resources and limiting data leakage.