SN 1093: Tokens in the Stream - Why LLMs are inherently insecure and prompt injection will persist

2026-08-26 · 168 min · episode 1093 · 15 entities

episode page audio

Asserted relationships

→ appeared on Security Now (Audio podcast

0.82

evidence rules-v5
Hosts: Steve Gibson and Leo Laporte
→ hosted by Leo Laporte person

0.55

evidence rules-v5
Feed author/publisher: Leo Laporte
→ discusses Technology company

0.40

evidence rules-v5
Feed category: Technology
→ hosted by TWiT company

0.40

evidence rules-v5
Feed author/publisher: TWiT
→ discusses Security concept

0.40

evidence rules-v5
Feed category: Security
→ references Clubtwit website

0.38

evidence rules-v5
Link in episode "SN 1093: Tokens in the Stream - Why LLMs are inherently insecure and prompt injection will persist": https://twit.tv/clubtwit
→ references SN 1093 Notes website

0.38

evidence rules-v5
Link in episode "SN 1093: Tokens in the Stream - Why LLMs are inherently insecure and prompt injection will persist": https://grc.com/sn/SN-1093-Notes.pdf

Entities found in this episode

companys 4

mentioned TWiT company

0.70

evidence rules-v5
Feed author/publisher: TWiT
mentioned Technology company

0.50

evidence rules-v5
Feed category: Technology
discusses Technology company

0.40

evidence rules-v5
Feed category: Technology
hosted by TWiT company

0.40

evidence rules-v5
Feed author/publisher: TWiT

websites 4

mentioned Clubtwit website

0.45

evidence rules-v5
Link in episode "SN 1093: Tokens in the Stream - Why LLMs are inherently insecure and prompt injection will persist": https://twit.tv/clubtwit
mentioned SN 1093 Notes website

0.45

evidence rules-v5
Link in episode "SN 1093: Tokens in the Stream - Why LLMs are inherently insecure and prompt injection will persist": https://grc.com/sn/SN-1093-Notes.pdf
references Clubtwit website

0.38

evidence rules-v5
Link in episode "SN 1093: Tokens in the Stream - Why LLMs are inherently insecure and prompt injection will persist": https://twit.tv/clubtwit
references SN 1093 Notes website

0.38

evidence rules-v5
Link in episode "SN 1093: Tokens in the Stream - Why LLMs are inherently insecure and prompt injection will persist": https://grc.com/sn/SN-1093-Notes.pdf

persons 3

mentioned Steve Gibson and Le person

0.86

evidence rules-v5
Hosts: Steve Gibson and Leo Laporte
mentioned Leo Laporte person

0.70

evidence rules-v5
Feed author/publisher: Leo Laporte
hosted by Leo Laporte person

0.55

evidence rules-v5
Feed author/publisher: Leo Laporte

podcasts 2

mentioned Security Now (Audio podcast

0.95

evidence rules-v5
Feed title: Security Now (Audio)
appeared on Security Now (Audio podcast

0.82

evidence rules-v5
Hosts: Steve Gibson and Leo Laporte

concepts 2

mentioned Security concept

0.50

evidence rules-v5
Feed category: Security
discusses Security concept

0.40

evidence rules-v5
Feed category: Security

Episode description as stored

Turns out, every chatbot conversation runs on a messy hack at the heart of language models, making prompt injection an unsolved—and possibly unsolvable—security threat. Steve and Leo unravel the research that explains why "roles" in AI aren't what you think they are. Understanding the controversy surrounding "AI Model Distillation" Anthropic moves to make their most powerful Mythos 5 model more widely available. Bitwarden's "Secrets Manager" offering prevents agentic and prompt injection abuse. The astounding and disturbing truth about the way conversation AI actually works Show Notes - https://www.grc.com/sn/SN-1093-Notes.pdf Hosts: Steve Gibson and Leo Laporte Download or subscribe to Security Now at https://twit.tv/shows/security-now . You can submit a question to Security Now at the GRC Feedback Page . For 16kbps versions, transcripts, and notes (including fixes), visit Steve's site: grc.com , also the home of the best disk maintenance and recovery utility ever written Spinrite 6 . Join Club TWiT for Ad-Free Podcasts! Support what you love and get ad-free audio and video feeds, a members-only Discord, and exclusive content. Join today: https://twit.tv/clubtwit Sponsors: threatlocker.com/twit box.com/AI hoxhunt.com/securitynow guardsquare.com material.security