Analysis
Prompt injection is not SQL injection (it may be worse) | National Cyber Security Centre
Publication date unknown · Discovered ncsc.gov.uk
Page published
Publication date unknown · First observed: 23 Oct 2026
Coverage timeline
Single-source analysis — one report is available.
Why it matters
Prompt injection can turn retrieved content into attacker-controlled instructions, so defenders must limit what assistants can access and execute rather than rely on filtering alone.
SUPERWISE’s reference guide explains direct, indirect, and triggered prompt injection and recommends layered defenses including least-privilege access, human approval, filtering, output checks, and logging. Supporting NCSC analysis explains why LLMs’ lack of a robust boundary between instructions and untrusted content makes comparisons with SQL injection misleading.