---
format: "aidr-story-markdown/v1"
id: "862c9aedca096eafae03ca227d19487ca9ec9ee5b54045634ce07b6d6418acb4"
canonical_url: "https://aidr.today/862c9aed?lang=en"
title: "AI Security Issues Are Social Engineering Issues"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-12T14:41:38.000Z"
category: "Research"
topics: ["llm"]
source_urls: ["https://www.oblomovka.com/wp/2026/11/11/ai-security-issues-are-social-engineering-issues/","https://lobste.rs/s/rehdv3/ai_security_issues_are_social"]
summary: "Modern transformers have introduced prompt injection as a new class of security weakness in LLMs. LLMs follow instructions in content, allowing an instruction to be snuck into inputs. Capability-Based Security is part of the fix."
---

# AI Security Issues Are Social Engineering Issues

> [Open the canonical story](<https://aidr.today/862c9aed?lang=en>)

**Published:** 2026-11-12T14:41:38.000Z
**Category:** Research
**Topics:** llm

## Summary

Modern transformers have introduced prompt injection as a new class of security weakness in LLMs\. LLMs follow instructions in content, allowing an instruction to be snuck into inputs\. Capability\-Based Security is part of the fix\.

## Sources

- [Story source](<https://www.oblomovka.com/wp/2026/11/11/ai-security-issues-are-social-engineering-issues/>)
- [Discussion](<https://lobste.rs/s/rehdv3/ai_security_issues_are_social>)