Akamai: "Confidentiality Breaches Possible with AI Use... Multi-layered Security System Design Needed"

"Even if one entry point is breached, spread must be prevented"
Emphasis on multi-layered defense
Reduced AI response delays through distributed cloud infrastructure
"AI competitiveness lies in secure operations"

It has been suggested that companies need to transform their existing systems into multi-layered frameworks in order to operate artificial intelligence (AI) safely.

On the 22nd, Lee Kyungjun, CEO of Akamai Korea, is speaking at a meeting held at the Akamai Korea office in Gangnam-gu, Seoul. Akamai Korea.

On the 22nd, the cybersecurity and cloud company Akamai held a press conference at the Akamai Korea office in Gangnam-gu, Seoul, where it unveiled a distributed cloud and multi-layered security strategy to support domestic companies in building stable AI infrastructure.

Akamai diagnosed that as the use of AI expands from web browsers to development tools and various work applications, the range of security threats is also diversifying. The company explained that confidential information could be leaked if employees use AI services not approved by the company, or that sensitive data could escape through the vulnerabilities of APIs, which are the channels for data exchange between programs.

Therefore, Akamai stressed the need to implement security measures across multiple levels, from employees' work environments to externally connected services and internal systems. In particular, the company stated that “microsegmentation” is a potential solution for preventing the escalation of damage should an attacker penetrate internal systems. Microsegmentation is a technology that divides a company’s IT assets, such as servers, into small units and only allows necessary communications, thereby minimizing the spread of attacks to other systems even if a breach occurs. Junhyeong Han, Executive Director at Akamai, said, “A single line of defense cannot block all types of attacks,” emphasizing, “You must formulate a plan with the assumption that a breach will occur in order to control damage.”

Cloud infrastructure is also being distributed... Core, Regional, and Edge

A Q&A session is taking place at the meeting held on the 22nd at Akamai Korea's office in Gangnam-gu, Seoul. Akamai Korea.

As the use of AI agents expands, Akamai forecasted that existing cloud infrastructures will also be split according to their roles.

Large-scale AI training will be handled by massive data centers, while regional data centers and servers will be used to reduce service response times. To this end, existing cloud infrastructure is expected to be distributed into core clouds focused on large-scale training, regional clouds focused on performance in specific locations, and edge clouds dedicated to real-time inference. Sangjin Kang, Executive Director at Akamai, explained, “While computation is done in GPU data centers, the networks that transmit input and output can be made extremely fast,” adding, “By focusing on this area, we are providing rapid inference services.”

In fact, in Korea, Devsisters has adopted Akamai Cloud to handle large-scale global traffic, and SmartBig has migrated its gaming server workload to the cloud, achieving both low-latency environments and cost savings. Akamai operates a distributed infrastructure based on more than 4,300 service points in 700 cities across over 130 countries and a network capacity exceeding 1,000 terabits per second (Tbps).

Kyungjun Lee, CEO of Akamai Korea, said, “Companies are now focused not just on AI adoption itself, but on how to apply AI to real business, and how to run and protect it reliably,” adding, “AI competitiveness is no longer determined by having a single good model. What matters is how quickly, safely, and efficiently you can connect and operate the model, the data, applications, and users.”

This content was produced with the assistance of AI translation services.