Cisco has announced four priority focus areas that organizations should consider to strengthen AI security as artificial intelligence adoption scales across industries in the Middle East.

The company reported that as AI moves from pilot projects to full production, security teams are under growing pressure to protect AI applications across the entire lifecycle. This includes securing data sources, models, and third-party components used in development and deployment.

According to Cisco, AI adoption is expanding rapidly across government, financial services, energy, and critical infrastructure sectors in the region. As a result, CISOs and IT leaders must adapt traditional application security practices to address emerging AI-specific risks and maintain digital trust.

Cisco revealed that AI application development often relies on open-source models, public datasets, and third-party libraries. These components can introduce vulnerabilities or malicious code that may compromise systems if not properly scanned and monitored.

The company also highlighted the importance of vulnerability testing. Static testing helps validate binaries, datasets, and models to identify risks such as poisoned data or backdoors. Meanwhile, dynamic testing evaluates how AI models behave in real-world production environments. Algorithmic red teaming can further simulate adversarial attacks at scale.

In addition, Cisco pointed to the growing role of AI application firewalls. These model-agnostic tools inspect AI traffic in transit and enforce security policies. They are designed to mitigate threats such as prompt injection, personally identifiable information (PII) leakage, and denial-of-service attacks.

Data loss prevention was also emphasized. Cisco noted that traditional DLP tools are less effective for AI-driven systems due to the dynamic nature of natural language content. Instead, AI-focused DLP examines both inputs and outputs to prevent sensitive data exposure.

Key AI security measures outlined include:

• Open-source scanning and vulnerability testing across AI components
• AI firewalls and data loss prevention to reduce risks like PII leakage

Commenting on the announcement, Fady Younes, Managing Director for Cybersecurity at Cisco Middle East, Africa, Türkiye, Romania and CIS, said organizations must secure AI applications beyond traditional controls. He added that applying established security principles in AI-specific ways can help organizations scale innovation while reducing risks such as prompt injection and sensitive data leakage.

Cisco concluded that risks exist at every stage of the AI lifecycle, from sourcing supply chain components to deployment. The company stated that a comprehensive AI security strategy is essential to protect applications from development through production.

Related post

View all