Analysis
Prompt injection is not SQL injection (it may be worse) | National Cyber Security Centre
First reported · Discovered ncsc.gov.uk
Page published
Earliest dated coverage: 29 Oct 2026 · First observed: 10 Nov 2026 · Latest dated coverage: 29 Oct 2026
Coverage timeline
Single-source analysis — one report is available.
Why it matters
The NCSC’s analysis highlights why defenders must account for untrusted content becoming instructions rather than assume LLM applications offer the instruction–data separation available in traditional systems.
The NCSC explains why prompt injection should not be treated like SQL injection: current LLMs do not enforce a security boundary between application instructions and untrusted content. Its analysis illustrates how attacker-controlled documents can redirect model behavior and warns that assumptions borrowed from SQL-injection defenses can undermine mitigations.