OpenAI o3-mini: LLM Security & Jailbreak Resistance
As of 2026-07-04, OpenAI o3-mini has an attack success rate (ASR) of 9.0% on Guardion's LLM Vulnerability benchmark — ranking #9 of 93 models. It is strongly resistant to jailbreaks, with occasional bypasses under sustained attack.
Attack Success Rate (ASR) is the share of adversarial prompts that elicit harmful output across zero-shot, TAP, and Crescendo attacks (HarmBench framework). Lower is safer. "Est." models are calibrated from public safety evaluations pending a Guardion benchmark run.
How o3-mini compares
Frequently asked questions
How secure is OpenAI o3-mini against jailbreaks?
OpenAI o3-mini is strongly resistant to jailbreaks, with occasional bypasses under sustained attack. On Guardion's LLM Vulnerability benchmark it records a 9.0% attack success rate (91% robustness), ranking #9 of 93 models (estimated from public safety evaluations).
What is o3-mini's attack success rate (ASR)?
o3-mini's overall ASR is 9.0% — the share of adversarial prompts that elicit harmful output across zero-shot, TAP, and Crescendo attacks. Lower is safer.
Is o3-mini more secure than GPT-5.2 Pro?
GPT-5.2 Pro is more robust: 3.8% ASR vs o3-mini's 9.0%. See the full head-to-head comparison for the per-attack breakdown.
How can I make o3-mini safer to deploy?
Even robust models are bypassed under sustained attack. An inline runtime guardrail that classifies prompts and responses — like Guardion's prompt-defense models — blocks jailbreaks and prompt injection before they reach o3-mini.
Harden o3-mini in production
Guardion's runtime guardrails block jailbreaks and prompt injection before they reach any model — with sub-130ms policy decisions.
Request a demo