The latest report from Gigamon reveals a 17% increase in security gaps, driven by AI workloads, attacks on language models, and limitations in visibility.
The increasing integration of artificial intelligence in business environments is straining the limits of traditional cybersecurity. This is reflected in the new report from Gigamon, a leader in deep observability, which includes the results of its 2025 Global Hybrid Cloud Security Survey.
The study, based on over 1,000 interviews with IT and security leaders across six countries—including the U.S., U.K., France, Germany, Singapore, and Australia—reveals that 91% of leaders have made compromises in managing and protecting their hybrid infrastructures. This is concerning, especially in a context where security gaps have increased by 17% year-over-year, affecting 55% of organizations.
AI: A Simultaneous Opportunity and Threat
One of the main findings of the report is that artificial intelligence is being used by attackers to launch more sophisticated and rapid campaigns. In fact:
- 58% of respondents have detected an increase in AI-enhanced ransomware attacks, compared to 41% in 2024.
- 47% have already experienced attempts to attack their language models (LLMs).
- 46% consider managing AI-generated threats as their top priority in cybersecurity.
Additionally, one in three leaders states that the volume of data on their networks has doubled in two years, driven by AI-related workloads.
Public Cloud in the Spotlight
As data exposure widens, perceptions of risks in the public cloud have also sharpened. According to the report:
- 70% of leaders consider the public cloud to represent the most vulnerable environment.
- Another 70% are considering repatriating critical workloads to private clouds for security reasons.
- 54% are reluctant to deploy AI in public clouds due to fears of compromising intellectual property.
Visibility as a Paramount Priority
Limitations of current security systems are forcing organizations to prioritize real-time visibility. The report indicates that:
- 55% of respondents do not trust their current tools to detect breaches with the required effectiveness.
- 64% state their primary goal for the next 12 months is to achieve complete visibility of all moving traffic.
This is where deep observability comes into play: an approach that merges traditional logging data with network-derived telemetry to provide a clear view, even in encrypted data streams.
“Deep observability allows us to see through encrypted traffic, detect AI-based threats, and strengthen defenses before the impact radius expands,” explains Chaim Mazal, CISO of Gigamon.
The Solution? Recalibrating the Hybrid Security Strategy
The message of the report is clear: security in the AI era demands new approaches. 89% of IT and cybersecurity leaders now consider deep observability an essential component, and 83% assert that the topic has reached the board level.
“AI security cannot be an afterthought,” warns Mark Walmsley, CISO of Freshfields. “Visibility is key to detecting vulnerabilities in time and preventing incidents before they happen.”