Skip to content

AI-Driven SOCs: Exaforce & Orange Cyberdefense Push LLM Limits

AI is transforming SOCs, but current Large Language Models struggle with real-time, high-precision tasks. Two companies are innovating to bridge this gap.

In the image in the center we can see one camera. On camera,it is written as "Sony Cyber Shot".
In the image in the center we can see one camera. On camera,it is written as "Sony Cyber Shot".

AI-Driven SOCs: Exaforce & Orange Cyberdefense Push LLM Limits

The shift towards AI-driven Security Operations Centers (SOCs) is pushing the boundaries of current Large Language Models (LLMs). While LLMs excel in everyday tasks, they face challenges in high-precision, real-time SOC environments. Two companies, Exaforce and Orange Cyberdefense, are working on solutions to overcome these limitations.

Exaforce is developing an evolved AI model that reduces human oversight in agentic systems. Meanwhile, Orange Cyberdefense is creating a solution for modern SOCs that consolidates platforms, automates processes, and filters noise. This allows analysts to focus on real threats, shifting teams from reactive to strategic roles.

In an AI-driven SOC, detection engineers focus on threat modeling and feedback loops. SOC analysts' productivity increases with automated triage bots. Threat hunters can proactively defend using real-time, context-rich information. The real-time data warehouse can also serve as a long-term warehouse for visibility and forensics, replacing legacy SIEMs. However, to achieve this, a new type of model that enables LLMs for the SOC use case is needed.

Humans excel at framing ambiguous problems and making risk-aware judgments, especially when signals are weak or novel. To shift towards machine-speed defense, LLMs must overcome limits such as real-time ingestion at scale, large durable context, low-latency execution, deterministic logic, and consistency of reasoning. With the efforts of companies like Exaforce and Orange Cyberdefense, the future of AI-driven SOCs looks promising, with machines excelling at processing high-volume, high-velocity, unstructured data.

Read also:

Latest