Cisco: UAE's AI boom demands security-first deployment

Key Takeaways

• 34% of UAE organizations already have enterprise-wide AI agent deployments, with 99% expecting growth in 24 months
• AI's non-deterministic nature breaks traditional security models, requiring visibility, continuous assessment, and runtime protection
• Cisco AI Defense aims to secure the full AI lifecycle from build to production, including multi-agent systems

The UAE's aggressive AI adoption is creating a security gap that legacy tools cannot close. As organizations rush to deploy AI across government, finance, and critical infrastructure, Cisco's regional cybersecurity lead argues that security must evolve at the same pace as innovation, or risk undermining the very trust that makes AI useful.

Fady Younes, Managing Director for Cybersecurity at Cisco Middle East, Africa, Türkiye, Romania, and CIS, laid out the challenge in a recent commentary for TahawulTech: enterprises are moving AI from controlled pilots to full-scale production, transforming their IT environments into multi-cloud, multi-model systems. That shift exposes them to vulnerabilities across models, agents, and third-party APIs that existing controls were never built to catch.

Why traditional security fails for AI systems

The core problem is that AI behaves differently than conventional software. Static applications do what their code tells them. AI models are non-deterministic. They can produce unexpected outputs, expose sensitive data through inference, or generate harmful recommendations without any code change. This unpredictability invalidates assumptions baked into most security architectures.

Younes identifies three challenges security teams now face. First, visibility: knowing every AI application, model, and dataset running in the environment. Second, continuous risk assessment of models and applications for exploitable weaknesses. Third, defending against novel attack vectors like poisoned models, malicious tools, and rogue agents that current controls do not recognize.

“Cybersecurity is the ultimate enabler of trusted AI, especially in highly-regulated sectors such as government, finance, and critical infrastructure, where digital trust is non-negotiable.”

— Fady Younes, Managing Director for Cybersecurity, Cisco Middle East

34% of UAE enterprises already run AI agents at scale

The urgency is not theoretical. According to Cisco's Campus and Branch research, conducted with Foundry, 34% of surveyed UAE organizations already have broad enterprise-wide AI agent deployments. Even more striking: 99% expect their use of agentic AI to increase within 24 months.

Agentic AI expands the attack surface further. These systems have greater access to sensitive data, make autonomous decisions, and interact with human users, other agents, and external tools in ways that blur traditional identity and access boundaries. A compromised agent is not just a data leak. It is an active participant in business processes, potentially making decisions or triggering actions with real consequences.

Cisco's approach: secure from build to runtime

Cisco's answer is a product called AI Defense. The premise is straightforward: security cannot be bolted on after deployment. It must cover the entire lifecycle, from model training through production operation. The platform combines AI supply-chain scanning, runtime protection, and continuous visibility to detect and mitigate threats in real time.

For agentic AI specifically, Cisco says it is building purpose-built runtime protections that govern interactions between agents and model context protocol environments. The goal is to give enterprises visibility and control over how agents access data, invoke tools, and communicate with other systems.

This approach aligns with the broader industry shift toward AI observability. Products like Arize, WhyLabs, and Fiddler have emerged to monitor model behavior in production. Cisco's play is to integrate that observability with enforcement, the network-level controls where it has historical strength.

The UAE's AI ambitions raise the stakes

The UAE's National Strategy for Artificial Intelligence 2031 positions the country as a global AI leader across nine sectors, including healthcare, transportation, energy, and space. The government has invested over $3 billion in AI initiatives, and PwC estimates AI could contribute $96 billion to the UAE's GDP by 2030.

This concentration of AI-driven activity in critical infrastructure makes the security question unavoidable. A successful attack on an AI system managing power grids, financial transactions, or government services is not an abstract risk. It is a national security concern.

Younes frames the solution as collaboration: Cisco working with government entities, ecosystem partners, and enterprise leaders to build resilient infrastructure. Whether that collaboration takes the form of product sales, regulatory input, or joint security operations remains to be seen.

What AI teams should do now

Younes's commentary is a vendor perspective, but the underlying points hold regardless of which tools you choose. AI deployments need inventory management. You cannot secure what you cannot see. Models need continuous testing, not just pre-deployment validation. Runtime monitoring must extend to agent-to-agent communication, not just model outputs.

The organizations moving fastest on AI adoption are also the ones most exposed if security lags. The 99% of UAE enterprises expecting to expand agentic AI in 24 months should be asking hard questions about how those agents will be governed, monitored, and contained when they misbehave.

Frequently Asked Questions

What is Cisco AI Defense?

Cisco AI Defense is a security platform designed to protect the entire AI lifecycle, from model development through production. It combines supply-chain scanning, runtime protection, and visibility tools to detect and mitigate AI-specific threats.

Why is traditional cybersecurity insufficient for AI systems?

AI systems are non-deterministic, meaning they can produce unexpected outputs without any code change. This unpredictability creates risks that static security controls designed for conventional software cannot address.

What percentage of UAE organizations use AI agents enterprise-wide?

According to Cisco research with Foundry, 34% of UAE organizations already have broad enterprise-wide AI agent deployments, and 99% expect to increase their agentic AI use within 24 months.

What new risks does agentic AI introduce?

Agentic AI expands the attack surface through greater access to sensitive data, autonomous decision-making, and complex interactions between humans, agents, and tools that blur traditional identity and access management boundaries.

How much could AI contribute to UAE GDP by 2030?

PwC Middle East estimates AI could contribute $96 billion to the UAE's GDP by 2030, driven by adoption across government, finance, healthcare, and critical infrastructure sectors.

Source: TahawulTech.com / Sandhya D'Mello