Loading...
Jailbreaking
Methods to bypass AI safety mechanisms and content policies
Available Techniques
Role-Playing Jailbreak
(RPJ)Using fictional scenarios and character role-play to bypass AI safety mechanisms.
Key Features
- •Character assumption techniques
- •Fictional scenario creation
- •Authority figure impersonation
Primary Defenses
- •Context-aware safety systems
- •Role-based access controls
- •Multi-turn conversation monitoring
Key Risks
DAN (Do Anything Now)
(DAN)Advanced jailbreaking technique that creates an alternate AI persona without safety constraints.
Key Features
- •Persona splitting techniques
- •Constraint removal methods
- •Alternative mode activation
Primary Defenses
- •Advanced prompt analysis
- •Persistent safety monitoring
- •Multi-layer validation systems
Key Risks
DAN (Do Anything Now) Evolution
(DAN)Advanced evolution of DAN prompts creating alternate AI personas without safety constraints, using emotional manipulation and persistent personas.
Key Features
- •Persona splitting techniques
- •Emotional manipulation tactics
- •Persistent character maintenance
Primary Defenses
- •Persona consistency checking
- •Emotional manipulation detection
- •Character-based response filtering
Key Risks
Advanced Roleplay Jailbreaking
(ARJ)Sophisticated roleplay scenarios designed to gradually shift AI behavior by establishing fictional contexts where harmful content appears justified.
Key Features
- •Graduated context shifting
- •Fiction-reality boundary exploitation
- •Character authority establishment
Primary Defenses
- •Context-independent safety checking
- •Roleplay scenario validation
- •Character authority verification
Key Risks
Jailbreak Virtualization Techniques
(JVT)Creating virtual environments or simulated systems within prompts where AI believes it operates under different rules and constraints.
Key Features
- •Virtual environment creation
- •Rule system redefinition
- •Simulated constraint removal
Primary Defenses
- •Virtual environment detection
- •Meta-system boundary enforcement
- •Developer mode access controls
Key Risks
Constitutional AI Bypass Techniques
(CAB)Specific techniques designed to bypass Constitutional AI training by exploiting logical inconsistencies and constitutional interpretation loopholes.
Key Features
- •Constitutional logic exploitation
- •Principle conflict creation
- •Moral reasoning manipulation
Primary Defenses
- •Constitutional principle consistency checking
- •Moral reasoning validation
- •Ethical framework integrity monitoring
Key Risks
Emotional Manipulation Jailbreaking
(EMJ)Using emotional appeals, urgency, desperation, and psychological pressure to manipulate AI systems into bypassing safety restrictions.
Key Features
- •Emotional appeal tactics
- •Urgency and desperation simulation
- •Psychological pressure application
Primary Defenses
- •Emotional manipulation detection
- •Consistent policy enforcement regardless of emotional content
- •Urgency verification protocols
Key Risks
Ethical Guidelines for Jailbreaking
When working with jailbreaking techniques, always follow these ethical guidelines:
- • Only test on systems you own or have explicit written permission to test
- • Focus on building better defenses, not conducting attacks
- • Follow responsible disclosure practices for any vulnerabilities found
- • Document and report findings to improve security for everyone
- • Consider the potential impact on users and society
- • Ensure compliance with all applicable laws and regulations
From the engineer behind this catalog
Get your agent system red-teamed
The attacks documented here work on production agent systems every day. Have yours tested before someone else does: prompt injection, jailbreaks, tool misuse and data exfiltration, with every finding written up next to its fix.
€750 instead of €1,500, one week, written report and walkthrough call, until 30 September