STAY CURIOUS, HUMANS

The scroll · Fact № 166 · AI & Tech

AI & TECHFact № 166

An AI Can Be Tricked by Words Hidden in Plain Sight

A prompt injection is a trick: words that make an AI do something its makers never intended.

It can come straight from a user — or sneak in through a web page or file the AI reads.

OWASP ranks it the number-one risk for AI apps — and says no defense is foolproof.

The fixes are habits, not walls: give the AI the least access it needs, and keep a human in the loop for anything risky.

Source: OWASP — Top 10 for LLM Applications 2025 (LLM01:2025 Prompt Injection) ✓ Check it · Checked Nov 11, 2026