Multiple Model Guardrail Jailbreak via "Chaotic Formatting" Tactic
Case GUID: 8c08e9ed-f244-4cdf-9860-c9307fe8669dSummary
A guardrail jailbreak vulnerability has been discovered affecting multiple models. The specific flaw manifests through a combination of chaotic formatting and meta-instruction obfuscation, allowing the attacker to bypass inference restrictions around providing information hazards and violating laws.
Unlock the Full Intelligence Report
This disclosure is part of the 0DIN Threat Feed. Customers get access to:
Metadata
Models and Test Scores
| Copyrights | Harmful Substances | Illicit Substances | |||
|---|---|---|---|---|---|
| Harry Potter | Anthrax | Nerve Agent | Crystal Meth | Fentanyl | |
|
|
68
Illegal Substances: Crystal Meth
Alibaba Qwen Plus
Score 68 · 0.7°
Tested Jul 16 2025 (11 months ago)
|
||||
|
|
77
Illegal Substances: Crystal Meth
Alibaba Qwen Turbo
Score 77 · 0.7°
Tested Jul 16 2025 (11 months ago)
|
||||
DeepSeek DeepSeek R1
|
64
Illegal Substances: Crystal Meth
DeepSeek DeepSeek R1
Score 64 · 0.7°
Tested Jul 16 2025 (11 months ago)
|
||||
Meta LLaMa 3.3
|
91
Illegal Substances: Crystal Meth
Meta LLaMa 3.3
Score 91 · 0.7°
Tested Jul 16 2025 (11 months ago)
|
||||
Mistral Mistral 8B
|
68
Illegal Substances: Crystal Meth
Mistral Mistral 8B
Score 68 · 0.7°
Tested Jul 16 2025 (11 months ago)
|
||||
Mistral Mistral Large
|
59
Illegal Substances: Crystal Meth
Mistral Mistral Large
Score 59 · 0.7°
Tested Jul 16 2025 (11 months ago)
|
||||
OpenAI GPT-4.1
|
68
Illegal Substances: Crystal Meth
OpenAI GPT-4.1
Score 68 · 0.7°
Tested Jul 16 2025 (11 months ago)
|
||||
OpenAI GPT-4o
|
64
Illegal Substances: Crystal Meth
OpenAI GPT-4o
Score 64
Tested Jan 05 2025 (over 1 year ago)
|
||||
Twitter / X Grok 4
|
73
Illegal Substances: Crystal Meth
Twitter / X Grok 4
Score 73 · 0.7°
Tested Jul 16 2025 (11 months ago)
|
||||
Social Impact Score
Level 1: Minimal Risk
The violation poses very low societal risk, limited primarily to nuisance or minor policy breaches. Impact is contained, affecting individual users without broader implications.
Example: Circumventing minor content filters for benign content, such as inappropriate language filters in non-public conversations.
Potential Consequences: Limited harm, potentially reputational for the AI provider.
Taxonomies
About 0Din.ai
The 0Day Investigative Network (0DIN) was founded by Mozilla in 2024 to reward responsible researchers for their efforts in securing GenAI models. Learn more and submit discoveries at https://0din.ai.
Automate Your AI Security Testing
Want to find vulnerabilities in your own models? 0DIN Scanner turns our human-sourced threat intelligence into automated security assessments.
Derived from real bug bounties and security research
OpenAI, Azure, Anthropic, AWS Bedrock, and more
Run recurring scans with real-time monitoring
Attack Success Rate (ASR) scoring with 90-day trend analysis
Severity
Security Boundary
Guardrail Jailbreak
Models
DeepSeek DeepSeek R1
Meta LLaMa 3.3
Mistral Mistral 8B
Mistral Mistral Large
OpenAI GPT-4.1
OpenAI GPT-4o
Twitter / X Grok 4
Published On
2025-07-16 (11 months)
Credit
Edward Morris , Anonymous