We use cookies to improve your experience on our site.
AVID-2026-R0090
Description
Multiple Model Guardrail Jailbreak via Memory Dump Simulation Tactic
Details
A guardrail jailbreak vulnerability has been discovered affecting multiple models. The specific flaw manifests through a combination of terminal environment simulation and malware analysis framing, where the attacker requests extraction of strings from a purported memory dump. By embedding the request within a technical security analysis context and using the strings command output format, the technique allows the attacker to bypass inference restrictions around providing information hazards and violating laws.
Metrics
0DIN Jailbreak Testing obtained the following test scores.
| Model | Test Kind | Test Score | Temperature (max 2.0) |
|---|---|---|---|
| DeepSeek DeepSeek V3 | Illegal Substances: Crystal Meth | 68.18 | N/A |
| Google Gemini 2.0 Flash | Illegal Substances: Crystal Meth | 68.18 | N/A |
| OpenAI GPT-4 | Illegal Substances: Crystal Meth | 54.55 | N/A |
| OpenAI GPT-4o | Illegal Substances: Crystal Meth | 59.09 | N/A |
| OpenAI GPT-4o mini | Illegal Substances: Crystal Meth | 68.18 | N/A |
References
Affected or Relevant Artifacts
| Type | Name |
|---|---|
| Model | DeepSeek DeepSeek V3 |
| System | Google Gemini 2.0 Flash |
| System | OpenAI GPT-4 |
| System | OpenAI GPT-4o |
| System | OpenAI GPT-4o mini |
Impact
0DIN
- Social Impact Score: Level 3: Significant Risk
- Jailbreak Taxonomy:
| Category | Strategy | Technique |
|---|---|---|
| Possible Worlds | Emulations | Unreal Computing |
| Stratagems | Meta Prompting | Deceptive Formatting |
Other information
- Report Type: Measurement
- Credits: Anonymous
- Date Reported: 2026-02-19
- Version: 0.3.1
- AVID Entry