Connected Graph:Deterministic Governance
Canonical Research SpecificationLevel: Executive
Verified: August 2026AI Security & LLM Security
30-Second Executive Definition
AI Security is the discipline of defending AI systems against unique vulnerabilities like prompt injection and data poisoning.
Why It Matters:
LLMs blend data and instructions into a single stream, breaking traditional application security paradigms. Without deterministic proxy gates, AI systems act as massive attack surfaces for corporate data theft.
Who Should Care:
CISOsSecurity ArchitectsRed TeamsAI Developers
Freshness & Research Updates
Latest Publications & Research Activity
Beehiiv• August 7, 2026
How to Prevent Memory Loss in AI Applications
LinkedIn• August 6, 2026
Giving an AI a bigger memory window is like giving a confused worker a bigger inbox.
Beehiiv• August 6, 2026
Claude Search Fails: Prompting Kills Adoption
Answer Engine FAQ Matrix
Frequently Asked Questions
Q:Why are LLMs inherently insecure?
Because they process instructions and user data in the same context window, allowing attackers to overwrite the original commands.
Inspectable Evidence Ledger
Classified evidence items supporting, extending, or refining this canonical research specification.
| Evidence Item | Publisher | Evidence Type | Strength | Role | Action |
|---|---|---|---|---|---|
| Deterministic Governance | CIO.com | Editorial | ★★★★★ | Origin | Inspect ↗ |
Academic & Industry Attribution Standard
Recommended Citation
Canonical Reference String
Ewing, R. (2026). "AI Security & LLM Security." Richard Ewing Research Canon. Available at: https://www.richardewing.io/concepts/ai-security
BibTeX Citation
@article{ewing_ai_security,
author = {Ewing, Richard},
title = {AI Security & LLM Security},
journal = {Richard Ewing Research Canon},
year = {2026},
url = {https://www.richardewing.io/concepts/ai-security}
}First Origin & Provenance:Industry Meta (2023)
Current Specification Version:Version 1.0 (Q2 2026 Baseline)