Story perspectives

Expert Warns: LLMs Unreliable for Critical Applications

1/26/2026

Summary

Story summary

A study by Vishal Sikka and Varin Sikka argues that large language models (LLMs) cannot reliably handle tasks beyond low complexity and is not yet peer-reviewed.
Vishal Sikka, former SAP chief technology officer, warns against using LLMs for critical applications because of AI hallucinations.
Some industry leaders expect improvements to mitigate problems, while Sikka says LLMs have limits and supplementary components could improve functionality, urging cautious AI integration.

1 of 1

Story summary

A study by Vishal Sikka and Varin Sikka argues that large language models (LLMs) cannot reliably handle tasks beyond low complexity and is not yet peer-reviewed.
Vishal Sikka, former SAP chief technology officer, warns against using LLMs for critical applications because of AI hallucinations.
Some industry leaders expect improvements to mitigate problems, while Sikka says LLMs have limits and supplementary components could improve functionality, urging cautious AI integration.