One AI attacker, 2,500 targets: why defences need validating at machine speed
Attackers no longer need to choose their targets one at a time. With AI handling the work, a single operator can run thousands of intrusions in parallel, and move from first access to data theft in minutes. In this Cyber Pulse interview, Tarek Kuzbari, VP Sales (EMEA) at Picus Security, explains what that means for the Middle East, and why security teams need to validate their defences at the same speed.
Key highlights
- CrowdStrike recorded a breakout time of just 27 seconds in 2025, and in some cases data theft began within four minutes of first access.
- In early 2026, one AI-assisted operator processed 2,516 targets across 106 countries in parallel and compromised more than 600 firewalls.
- Middle East breach costs dipped in 2025 but rose to an average of $8 million in IBM's 2026 findings.
- Picus Exposure Validation tests vulnerabilities against an organisation's own security controls, so teams fix what is actually exploitable.
- Tarek's message: "It's not about prediction, it's not about assumption. What we do is actual validation."
The speed gap is real
CrowdStrike's 2026 Global Threat Report put the average eCrime breakout time at 29 minutes, with the fastest recorded at 27 seconds. In some cases, data exfiltration began within four minutes of initial access. Against that pace, a security programme that tests its defences once a quarter is working on very old information.
Tarek is clear that the region is not on the sidelines.
"Practically the Middle East is one of the most targeted regions across the globe."
— Tarek Kuzbari, VP Sales (EMEA), Picus Security
IBM's Cost of a Data Breach research shows how volatile the cost can be. Its 2025 findings for Saudi Arabia and the UAE showed an 18% drop, to SAR 27 million per breach. Tarek expected that to reverse, given geopolitical tension and attackers' growing use of AI. IBM's 2026 findings put the regional average back up at $8 million.
One operator, thousands of targets
In February 2026, researchers uncovered a live AI-assisted campaign against FortiGate firewalls. According to Amazon Threat Intelligence and independent researchers, the operator got in through exposed management interfaces and reused passwords, then handed off to AI. A custom MCP server linked large language models, including DeepSeek and Claude, to the attack tooling, which mapped paths to Active Directory, escalated privileges and moved laterally. More than 600 devices in 55 countries were compromised, and the operator was processing 2,516 targets across 106 countries in parallel batches.
For Tarek, the scale is the real lesson. Organisations used to assume they were too small to be first in line, or that a patch would arrive before attackers reached them.
"These attackers, they were doing 2,500 attack[s] at the same time. Can they do 25,000 attack[s]? Yes, it's just an additional MCP."
— Tarek Kuzbari, VP Sales (EMEA), Picus Security
With attacks automated at that scale, every organisation needs to find its highest-impact gaps, across security controls and vulnerabilities, and close them fast.
Too many vulnerabilities, not enough time
The volume problem is just as hard. According to Picus, more than 40,000 new CVEs were disclosed in 2024, and 61% were labelled high or critical. No team can patch them all, and a CVSS score alone does not say whether a vulnerability can be exploited in a specific environment.
"It's not about prediction, it's not about assumption. What we do is actual validation."
— Tarek Kuzbari, VP Sales (EMEA), Picus Security
How Picus Exposure Validation works
Picus launched Exposure Validation in May 2025. Tarek describes the approach in four steps:
- Bring in findings from the organisation's vulnerability management tools, red team work and threat intelligence.
- Test whether existing security controls detect and stop attacks that use those vulnerabilities.
- Re-score each vulnerability based on whether it is actually exploitable in that environment, not just its CVSS score.
- Map it to the organisation's assets, so vulnerabilities affecting critical systems rise to the top.
The result is a shorter, evidence-based list. In one example Picus shared, 63% of a global industrial company's vulnerabilities were rated critical by CVSS alone, but only 9% proved truly high risk once tested.
Amiviz and Picus in the region
Amiviz has worked with Picus Security since 2021, bringing its security validation platform to resellers and customers across the Middle East.
What this means for MEA organisations
Attackers now test your defences continuously, at a scale no human team can match. Quarterly or annual testing leaves long windows where nobody knows whether controls still work. Continuous validation answers two practical questions at any moment: would our controls stop the attacks targeting the region right now, and which of our vulnerabilities can actually be exploited? That turns a list of thousands into a short list of fixes that matter.
Validate your defences before attackers do
Want to see which gaps in your environment attackers could actually exploit? Book a demo through our Picus Security page, email info@amiviz.com or contact us.