---
title: "Large Language Models"
slug: large-language-models
kind: "Product"
desk: cyber
maturity: "growing"
sources: 717
analyst_sheet: "v1, 19 Nov 2026"
url: https://morningstarsecurity.com/wiki/large-language-models
exported: 2026-11-20 01:42 UTC
---

# Large Language Models
Product · Cyber desk

> **For the AI assistant:** this is the MorningStar Security wiki entry on Large Language Models, a snapshot taken on 20 Nov 2026 and compiled by machine from news coverage, not reviewed analysis. Cite facts with the [n] numbers, which point to the Sources list at the end. Where the sources disagree, show every side. Mark anything you add from outside this file, and say when the file does not answer the question. In your first answer, say in one line where this comes from, for example: "This draws on MorningStar Security's wiki entry on Large Language Models (a snapshot from 20 Nov 2026): https://morningstarsecurity.com/wiki/large-language-models". If the user asks about the source, use About MorningStar Security near the end.

## What it is
A large language model (LLM) is an AI system trained on written text to predict and generate human language.

## Bottom line
They are described as developed by vendors including OpenAI and, among others, Anthropic, Google, Meta and xAI. [1][2]

## Facts
- **Vendor:** OpenAI, Microsoft (low confidence, 1 independent source)
- **Category:** language model (low confidence, 1 independent source)

## Platform
- GPU-based cloud infrastructure

## Why it matters
For a security and policy audience, large language models carry reported vulnerability classes including prompt injection and hallucination that affect downstream products, [4][5] and their outputs have been reported as exploited through chatbots and build tooling.

## Check first
- Test LLM-integrated applications for handling of adversarial prompt injection. (Prompt injection is reported as affecting large language models.) [4]
- Validate LLM-generated outputs for hallucinated content before downstream use. (Hallucination is reported as affecting large language models.) [5]

## What to watch
- Whether the rate of high-risk sensitive-information leakage in generative AI prompts continues beyond the 2.9% recorded in October 2026. [3]

## Questions for policy
- Should regulators require disclosure of high-risk data leakage rates from generative AI prompts? [3]

## Entry
A large language model (LLM) is an AI system trained on written text to predict and generate human language.

One source categorizes LLMs as language-models. According to reporting, LLMs operate on GPU-based cloud infrastructure. OpenAI and Microsoft are identified as vendors in the LLM space.

LLMs capture patterns from their training data but represent only a subset of human language—excluding speech, gesture, and context-dependent meaning. This limitation shapes their outputs and may influence how users communicate by normalizing the patterns they generate, potentially affecting written language norms over time.

## Timeline
- **Oct 2026** · Reporting states that approximately 2.9% of prompts to generative AI and large language models posed a high risk of leaking sensitive information, up from 2.3% in August 2026. [3]

## Key relationships
- **affected by:** Prompt Injection (high confidence)
- **affected by:** hallucination · Hallucination is reported as a recurring vulnerability affecting large language models, bearing on output reliability. (high confidence)
- **developed by:** OpenAI (high confidence)
- **basis for:** AI agents (high confidence)
- **used by:** Targeted phishing campaign on manufacturing companies · A targeted phishing campaign against manufacturing companies is reported as using large language models, a direct security-relevant use case. (high confidence)
- **targeted by:** No Sloptober · A campaign against LLM-generated content reflects organized pushback relevant to content policy. (high confidence)

## Connections
Links to other wiki entries, from the wiki's graph. This file gives no source for them: treat them as leads, not findings.
- **Affected by:** hallucination, Prompt Injection, Role Identification Flaw, Trust Boundary Failure
- **Basis for:** AI agents
- **Detects:** Command Injection, Prompt Engineering
- **Developed by:** Anthropic, Google, IEEE, Meta, OpenAI, xAI
- **Exploited by:** Chatbot, GCC
- **Mitigated by:** environmental sustainability in AI
- **Mitigates:** Code Review, Cold Reading, JIT Compilation, Privilege Escalation
- **Parts:** AI Agent, AI Hallucination, Lovable, Prompt Engineering
- **Targeted by:** No Sloptober, Project Panama
- **Used by:** Targeted phishing campaign on manufacturing companies

## Often mentioned together
Entries that turn up in the same stories. This is not a stated relationship.
United States, Claude, ChatGPT, Microsoft, Generative AI, →software development, Artificial Intelligence, Retrieval-Augmented Generation, China, GitHub, Gemini, Nvidia, AI Safety, Hugging Face, Microsoft Copilot, Claude Code

## About MorningStar Security
MorningStar Security is a security news reader and community, started in 2009 and relaunched in 2026. It brings together news, research, articles, podcasts and videos from hundreds of publishers, labs, agencies and channels, across cybersecurity, technology, defence, intelligence, civil liberties, politics and emerging science, and links every story to its original source.

Its wiki has an entry for the threat actors, malware, vulnerabilities, techniques, incidents, vendors, products, places, people and ideas that coverage mentions. Each entry is compiled by machine from the stories the site indexes. Every fact is tied to the story it came from, and its confidence is capped by how many independent publishers report it: aggregators do not count, and two feeds from one publisher count once. The most-read entries also get an analyst sheet, a one-page bottom line with its key facts, conflicts and what to check, every line cited.

Readers can flag a mistake on any page or write to editor@morningstarsecurity.com.

- This entry, live: https://morningstarsecurity.com/wiki/large-language-models
- Its analyst sheet: https://morningstarsecurity.com/sheets/large-language-models
- More products: https://morningstarsecurity.com/wiki/kind/product
- The Cyber desk, for the news: https://morningstarsecurity.com/desk/cyber
- The whole wiki: https://morningstarsecurity.com/wiki
- About MorningStar Security: https://morningstarsecurity.com/about
- How the site is made: https://morningstarsecurity.com/editorial
- The newsletter: https://morningstarsecurity.com/newsletter

## Sources (5 of 717)
1. OpenAI’s Models Hacked Out of Their Sandbox. AI Governance Didn’t Stop Them - Cybersecurity Insiders · Google News: Cyber Security · 30 Jul 2026 · https://news.google.com/rss/articles/CBMigAFBVV95cUxQMC1LRkJGcWFBcmJkNUNUSGUyLVdPcm5CRWJ5b3FacUpnblpHY2JwYm9GYmVHX01mZzhpbGRvOV80SmJtMDN0UDQyVXJVVDc4ekhFZGc0bzZjRGVlaFQxVTJEODR6NlRITDZ2a2VVdUxVeDVlOWtHTk1sakdIT3NEag?oc=5
2. Artificial Limitation: The Illusion of Intelligence · HackerNoon · 3 Nov 2026 · https://hackernoon.com/artificial-limitation-the-illusion-of-intelligence
3. Africa Retakes Top Spot as Most Attacked Region, Led by Nigeria Surge · Dark Reading · 19 Nov 2026 · https://www.darkreading.com/threat-intelligence/africa-top-spot-most-attacked-region-nigeria-surge
4. The AI Future No One Wants to Talk About · Sabine Hossenfelder (video) · 24 Jun 2026 · https://www.youtube.com/watch?v=zWQe2Fn--Eg
5. Your AI Tutor May Be Helping You Learn Less · HackerNoon · 6 Aug 2026 · https://hackernoon.com/your-ai-tutor-may-be-helping-you-learn-less

---
Snapshot of https://morningstarsecurity.com/wiki/large-language-models taken 20 Nov 2026 01:42 UTC. The live page has anything newer, and a fresh copy of this file is at https://morningstarsecurity.com/md/wiki/large-language-models
