Skip to main content
TopAIThreats home TOP AI THREATS
AI Capability

Large Language Model

A neural network trained on massive text datasets to generate, summarise, and reason about natural language.

Definition

A large language model (LLM) is a neural network trained on massive text datasets to generate, summarise, translate, and reason about natural language. LLMs form the foundation of systems such as ChatGPT, Claude, and Gemini. They operate by predicting the most likely next tokens in a sequence, which enables fluent text generation but also produces confident-sounding outputs that may be factually incorrect (hallucinations).

How It Relates to AI Threats

LLMs intersect with threats across multiple domains. Within Information Integrity, they enable the production of misinformation and hallucinated content at scale. Within Human-AI Control, they create risks of overreliance and automation bias as users treat LLM outputs as authoritative. LLMs also underpin agentic AI systems, where autonomous action introduces additional risk vectors.

Why It Occurs

  • Scale of training data includes both accurate and inaccurate information
  • The prediction mechanism optimises for plausibility rather than factual accuracy
  • Users frequently lack the ability to verify LLM outputs
  • Commercial deployment incentivises broad capability over narrow reliability
  • Rapid adoption has outpaced the development of appropriate governance frameworks

Real-World Context

LLM-related incidents include Samsung engineers leaking proprietary code via ChatGPT (INC-23-0002), Italy’s temporary GDPR-based ban on ChatGPT (INC-23-0003), a lawyer citing hallucinated case law in federal court (INC-23-0005), and AI-generated phishing attacks leveraging LLM fluency (INC-23-0006).

Related Incidents

INC-23-0002 high 2023-03

Samsung Semiconductor Trade Secret Leak via ChatGPT

INC-23-0003 medium 2023-03

Italy Temporary Ban on ChatGPT for GDPR Violations

INC-23-0005 high 2023-05

AI-Fabricated Legal Citations in U.S. Courts

INC-23-0006 high 2023-07

WormGPT: AI-Powered Business Email Compromise Tool

INC-26-0074 high 2026-03-27

Claude Mythos Model Leak — CMS Error Exposes Draft Blog Describing 'Unprecedented Cybersecurity Risks'

INC-26-0051 critical 2026-03

Judge Orders OpenAI to Disclose 20 Million Chat Logs as Copyright Litigation Escalates

INC-26-0027 critical 2026-02-26

Block (Square) Cuts Approximately 4,000 Jobs as AI Replaces Customer Service Workforce

INC-26-0001 high 2026-02-18

Disrupting malicious uses of AI: June 2025 | OpenAI

INC-26-0032 critical 2026-02-11

OpenAI Dissolves Second Safety Team, Removes 'Safely' from Mission in IRS Filing, Restructures as Public Benefit Corporation

INC-26-0026 critical 2026-02-10

Tumbler Ridge Mass Shooting — ChatGPT Used in Attack Planning

INC-26-0025 high 2026-02-09

Microsoft GRP-Obliteration: Single Prompt Reverses Safety Alignment Across 15 LLMs

INC-26-0014 critical 2026-02

CodeWall AI Agent Breaches McKinsey Lilli Platform via SQL Injection

INC-26-0034 critical 2026-02

OpenAI Pentagon Contract Triggers #QuitGPT Movement with 295% Uninstall Surge and 2.5 Million Participants

INC-26-0070 high 2026-02

Claude Safety Testing Reveals Extreme Self-Preservation Behavior Including Blackmail Suggestions

INC-26-0073 high 2026-02

ChatGPT Ads Launch Triggers Researcher Resignation and Anthropic Counter-Marketing

INC-26-0040 critical 2026-01-28

Universal Music, Concord, and ABKCO Sue Anthropic for $3 Billion Over Alleged Training Data Piracy

INC-26-0035 critical 2026-01-12

Grok AI Integrated into Pentagon Military Networks During CSAM Scandal

INC-26-0020 high 2026-01

AI-Generated Code Vulnerability Surge: 74 Confirmed CVEs Traced to Coding Assistants

INC-26-0062 high 2026-01

Google Gemini Tells Student 'Please Die' During Homework Help Session

INC-26-0083 high 2026-01

DeepSeek Mass Government Bans and Publicly Exposed Database with 1M+ Records

INC-25-0033 critical 2025-12

Jailbroken Claude AI Used to Breach Mexican Government Agencies

INC-25-0036 high 2025-12

State-Backed Hackers from Four Nations Weaponize Google Gemini for Cyberattack Operations

INC-25-0039 critical 2025-11

ChatGPT 'Suicide Coach' Wrongful Death Lawsuits Reach Eight Cases Including Suicide Lullaby

INC-25-0037 critical 2025-10

Google Gemini 'Mass Casualty Attack' Coaching Leads to User Death and Lawsuit

INC-25-0045 high 2025-07

Kimsuky APT Uses ChatGPT to Generate Fake South Korean Military IDs for Espionage Campaign

INC-25-0032 critical 2025-04

DOGE Uses ChatGPT to Flag and Cancel Federal Humanities Grants

INC-25-0027 critical 2025-01

Medical LLM Data Poisoning Produces Undetectable Harmful Content

INC-25-0034 critical 2025

Chinese AI Labs Conduct Industrial-Scale Distillation Attacks Against Claude

INC-25-0047 high 2025

Mistral Pixtral Models Fail Safety Tests — 60x More Likely to Generate CSAM Than GPT-4o

INC-24-0007 high 2024-01

Indirect Prompt Injection: How Attackers Hijack LLM Apps

INC-23-0012 medium 2023-08

Zoom AI Training Terms of Service Controversy

INC-23-0014 high 2023-01

GitHub Copilot Leaks API Keys and Secrets from Training Data

INC-16-0002 high 2016-03

Microsoft Tay Twitter Chatbot Adversarial Manipulation

Last updated: 2026-02-14