IBM and Akamai address security and identity risks of AI agents

Industry leaders are addressing the security and operational challenges posed by the rise of autonomous AI agents. At the Financial Tech Conference 2026, IBM executive Choi Seok-jae emphasized that controlling AI agents requires assigning them unique identities. He warned of “shadow actors”—uncontrolled AI agents that lack individual identities, making it difficult to track actions or assign accountability when agents perform unauthorized tasks, such as deleting databases or accessing sensitive personnel data.

To mitigate these risks, IBM suggests implementing “actor tokens” to provide dynamic, just-in-time permissions rather than permanent access. This approach aims to ensure that AI agents operate with the minimum necessary authority for specific tasks.

Separately, at Akamai’s 2026 Media Briefing, Akamai Korea CEO Lee Kyung-jun noted that the focus for enterprises has shifted from merely adopting AI to ensuring its stable operation and security. As AI agents gain autonomy, security measures must evolve to defend at “machine speed” to counter threats like “agent hijacking,” where malicious instructions are hidden within external documents or web pages.

Akamai is also pivoting its infrastructure focus toward the AI inference market. Rather than focusing solely on model training, the company is prioritizing low-latency inference infrastructure to support real-time services such as chatbots, recommendation systems, and gaming.