"UK Cyber Agency Warns of Potentially Fundamental Flaw in AI Technology"

Britain's National Cyber Security Centre (NCSC) has issued a warning about a fundamental security vulnerability impacting Large Language Models (LLMs), the type of Artificial Intelligence (AI) used by ChatGPT to perform human-like conversations. Since ChatGPT's launch in November 2022, most security concerns regarding the technology have centered on its ability to automatically generate human-like speech. Cybercriminals are now deploying their own versions to generate "remarkably persuasive" phishing emails. In addition to the malicious use of LLM software, there are vulnerabilities stemming from its use and integration with other systems, especially when the technology interfaces with databases and other product components. It is referred to as a "prompt injection" attack, and according to the NCSC, the issue may be fundamental. The agency warned that research suggests an LLM cannot distinguish between an instruction and data provided to help complete the instruction. This article continues to discuss Britain's NCSC warning of a fundamental flaw in AI technology. 

The Record reports "UK Cyber Agency Warns of Potentially Fundamental Flaw in AI Technology"