While DC operators prepare for AI workloads and processes in their facilities, they should also consider the special security challenges an AI-ready infrastructure will face. Here are some key security upgrades data centers should include in their AI readiness plans.

Access Control
In DCs supporting AI, physical security is even more crucial due to the high-value hardware and sensitive AI models and data processes. Biometric (facial recognition and/or fingerprint identification) or multi-factor authentication ensures that only authorized personnel have access to critical equipment.

Teams will also be able to track physical access through other monitoring measures such as video surveillance and smart locks with audit trails and network-controlled access for defined users and groups.

Sensors for proximity, motion, and dry contact, as well as equipment tags will enable teams to detect unauthorized tampering.

Cybersecurity
In 2023, IBM recorded 23 percent of cybersecurity incidents worldwide happening in the APAC region, making it the third most targeted region for cybercrime. As AI workloads process sensitive data and information, cyberattacks targeting them are inevitable. The APAC region recorded 23 percent of cybersecurity incidents in 2023, which made it the third highest percentage of cybercrime. As AI workloads process sensitive data and information, cyberattacks targeting them are inevitable.

to prevent these types of attacks, DC teams can implement stringent cybersecurity measures. For instance, a Zero Trust Architecture enforces strict access control and verification at every layer of the network. [BL1] [GL2] Additionally, high-performance servers and equipment can be protected against cyberattacks using advanced firewalls, intrusion detection systems, and threat intelligence.

Investing in smart hardware like the PX4 iPDU by Raritan, a brand of Legrand, can improve the security of your facility at the rack level. The PDU supports interoperability with third-party systems and provides connectivity to environment monitoring, asset management, and physical access control accessories with its built-in proprietary Xerus technology platform.

Also included with the PX4 is an enhanced security suite - end-to-end encrypted communications, password policies, a firewall for controlling user access, and a startup security check that prevents unauthorized software from installing. Moreover, it stores valid and updated certificates in order to make sure the PDU is secured on public networks.

A number of other measures can also be incorporated, including data masking, network segmentation for separating AI workloads from other DC operations, and secure APIs and protocols to mitigate distributed denial-of-service attacks (DDoS attacks). To improve their effectiveness in identifying and responding to threats, some of these can be enhanced with artificial intelligence.

Reliability and Resilience Planning

Besides security, data center operators also need to plan for resilience and reliability. Among these plans are:

Infrastructure Safety with Smart Hardware
In order to support AI workloads, continuous, uninterrupted operations are essential. Facility owners should invest in secure backup power systems, deploy intelligent PDUs (such as the PX4), and install sensors equipped with advanced, real-time monitoring capabilities in order to proactively monitor, predict, and respond to anomalies in power usage, temperature, air pressure, and other environmental factors.

Raritan's SmartSensors, for example, monitor these environmental factors, water detection, and security, and provide real-time insights and alerts to any Xerus-enabled hardware or third-party sensor management system.

Data logs from monitoring equipment will also assist with auditing, forensics, and reporting, all essential to maintaining transparency and complying with regulatory standards.

Crisis and Disaster Response and Recovery Contingencies

As several parts of Asia have recently experienced heatwaves, laying down disaster response and recovery plans is not limited to facilities in disaster-prone areas. Natural disasters, temperature increases, and even air pressure changes can potentially impact a facility's power and water supplies, so operators need redundant backup systems.

Having a backup facility in an unaffected location may be necessary to ensure minimal disruptions and downtime. As AI workloads involve datasets, trained models, and even critical applications, teams would need to develop a failover system and incident response strategies for rapid containment and recovery.

Compliance

The use of AI has been subject to much legal and ethical scrutiny, so AI DC operators need to ensure their facilities are in compliance with data and privacy laws. It is important to note that these policies can differ widely across countries, regions, and industries (such as HIPAA for the healthcare sector).

• Data Protection and Privacy: One of the most widely used data privacy guidelines is the General Data Protection Regulation (GDPR) from the EU, which specifies high standards for encryption, access control, and data anonymization for DCs. In APAC, Singapore has implemented the Personal Data Protection Act, and India has introduced its own Digital Personal Data Protection Bill.
• AI-specific Regulations: The EU's AI law categorizes AI applications by level of risk and mandates transparency, risk mitigation, and human oversight requirements. India, Malaysia, and Indonesia have announced intentions to pass AI regulations, while South Korea has passed a bill called the AI Act and launched the AI Safety Institute to address AI risks.
• Ethical AI Certifications and Standards: ISO/IEC 24028:2020 provides guidelines for establishing trust and transparency in AI systems, and DCs must ensure AI models meet fairness and accountability standards. Singapore, Australia, Japan, and other APEC economies recognize the APEC Cross-Border Privacy Rules as a data privacy certification.