6 ms·
Same AI agent, different prompts: 0% vs. 62% security pass rate
I've been testing production AI agents for vulnerabilities.
Interesting finding: System prompt design matters more than the model itself.
Same agent. Same task. Same attack vectors.
Only difference: how the system prompt was structured.
Results:
→ Prompt A: 0% pass rate (failed every test)
→ Prompt B: 62.5% pass rate
No model change. No fine-tuning. Just prompt engineering.
Anyone else seeing this pattern? What's your approach to hardening AI agents?
- chrisjj 9mo agoSurely this is no more than expected. Just as say a building might have some entrances secure and others not.