Glossary
LLM
Large Language Model — A type of AI model trained on vast amounts of text to understand and generate human-like language, underpinning tools like chatbots and coding assistants. LLMs introduce novel security risks distinct from traditional software, including prompt injection, data leakage through model outputs, and hallucinated but confidently stated false information.
Recent coverage mentioning LLM
GTIG AI Threat Tracker: Evolution of Adversarial Agentic AI
Google Threat Intelligence Group tracks threat actors shifting to agentic AI, targeting proprietary models, and abusing open source software.
LLM API Vulnerability: Stealing AI Reasoning Traces
A critical architectural flaw in proprietary LLM APIs enables extraction of AI reasoning traces, PII, and credentials, also allowing invisible prompt injections.
The AI Safety Penalty: How LLM Guardrails Hinder Defenders
Cisco Talos warns about the 'AI safety penalty,' where large language model guardrails impede legitimate defensive operations, giving attackers an advantage.
AI-Assisted Campaigns Target Latin American Organizations
Ongoing multi-stage network intrusions and data exfiltration campaigns in Latin America leverage AI tools to enhance operations.
Rogue LLM Endpoints: Data Exposure & RCE Risk for AI Agents
Unverified LLM endpoints pose significant risks, enabling data leakage and potential remote code execution via compromised AI agent sessions.
Diagnosing LLM Safety Fragility with Perturbation Probing
New research introduces Perturbation Probing to diagnose LLM safety fragility, revealing guardrails are often concentrated in few neurons.
Advertisement