AI Knowledge Poisoning
Planting false facts where AI systems and search engines will retrieve them — wikis, vector stores, web pages — so the assistant confidently repeats the attacker's claim; injected content can even become instructions.
An everyday example
A vendor comparison page seeded so that internal AI tools summarise a competitor as 'under regulatory investigation'.
What it can cost you
Trusting AI summaries as neutral; decisions built on poisoned retrieval; assistants that follow instructions hidden in content.
Warning signs
- The AI said so and I did not check.
- The claim exists only in summaries.
One thing to do
Treat AI output as a lead, not a source: trace material claims to primary documents; ask tools for citations and check them; keep retrieval corpora curated.
Is this happening to you right now?
Describe your situation and get a read across all six brains. No account needed.
Check my situationRelated
Not medical, legal or financial advice. How Decision IQ works
One decision idea a week
One useful idea. No daily drip, no pressure.