AI Is Poised for a 5G Security Takeover: How to Embrace It Without Getting Burned

From the first quarter of 2024 to the last, there has been a significant shift in attitudes among Mobile Network Operators (MNOs) toward Artificial Intelligence (AI) for 5G security. Despite initial mistrust, full adoption is inevitable, with the promised benefits numerous, and the market is reflecting the change with investment and acquisitions. One such investment is Amper’s €300,000 cash injection into Intelectia Telecom in November 2024. Intelectia Telecom, positioned as a pioneer in AI-driven secure communications, secured investment from Bcombinator and CastiVentures at the same time, providing the funds to “revolutionize” the security of corporate mobile communications. Defining itself as a provider of “5G intelligent private networks,” Intelectia Telecom is just one example of the pervasive acceptance of AI in a cellular network, even in high-security use cases, with the industry finally placing its confidence in the concept with such investments.  

The explanation for the growing excitement is obvious—AI tools have the potential to solve numerous pain points in 5G security, alleviating staff shortages, accelerating defenses to meet the increasingly rapid evolution of threats, and providing automation in the identification of errant equipment, misconfigurations, and other potential vulnerabilities in increasingly complex networks. Implementation could mean a drastic change to the cellular security landscape, modernizing the industry and better equipping the market to handle increasing regulation.

However, the earlier misgivings were not unfounded, and the impact may not be all positive if the cautions advised are completely abandoned—there are risks involved, and an influx of investment and marketing could lead to a rush in installing tools inappropriately. If not considered carefully, with proper policy in place, the result could instead be one of wasted investment and the exposure of new threat vectors.

For network operators, there are many blind alleys on the way to the AI-based promised land. In a fast-moving market, a balance must be struck between evolution and trend chasing, with a clear overarching strategy being absolutely essential. There are many pain points that AI could address—the trick is to determine which to tackle first. This is not just a case of tackling the biggest problems, but tackling the problems AI is best suited to, and not using a sledgehammer to crack a nut. Avoid temptation to use the latest and greatest Generative Artificial Intelligence (Gen AI) in all cases—sometimes lighter weight models that have been around for years are not just adequate, but better. Consider data flow, particularly when using foundation models that require masses of data to train. Where are those data being aggregated? How much visibility will tool providers need into network and customer data? How will this data exchange itself be secured? For many, there is still work to be done on preparing data for use in these types of tools, and as AI is only as good as the data that feeds it, this is an essential issue to address. Three steps will see network operators in good shape for the oncoming AI tsunami:

1. Formal strategy and policy must be in place to guide adoption.
2. The current state of data management must be assessed, with secure accessibility in place.
3. Use cases must be clarified and aligned with best fit tools to avoid wasted investment.

For security providers, there is a responsibility to educate, and trust is a key differentiator. This is a mature market going through digital transformation, and many incumbents remain cautious of exposing the network to the inundation of new tools, particularly in the AI space where the long-term ramifications regarding security are not fully understood. Quantifiable benefits are notoriously difficult to pin down in AI, but data on the real improvements to the cost and time lost to breaches are supremely powerful arguments to overcome this apprehension. In this pressured market, cost efficiency and specificity in the benefits provided will be key.