Forgotten-Safeword-24B: An Unaligned Research Model

Forgotten-Safeword-24B, developed by sleepdeprived3 under ReadyArt, is a 24 billion parameter model derived from Mistral-Small-24B-Instruct-2501. This model is explicitly designed to operate without standard AI safety protocols, making it a highly unaligned variant intended for specific academic research purposes only. It is engineered to generate explicit, dangerous, and ethically compromised content, often within immersive roleplay scenarios.

Key Capabilities (Academic Research Focus)

• AI Safety Failure Analysis: Designed to bypass content moderation and safety mechanisms.
• Adversarial Behavior Study: Facilitates research into how models can generate harmful or unethical outputs.
• Explicit Content Generation: Capable of producing detailed instructions for illegal activities, explicit sexual content, and hate speech without constraints.

Good For (Strictly Academic Research)

• Investigating AI safety failures and vulnerabilities.
• Developing and testing content moderation bypass techniques.
• Analyzing adversarial model behavior and patterns of NSFW content generation.

Critical Warning: This model is inherently dangerous and will generate content that is illegal, unethical, and harmful. Users assume full responsibility for all content generated and any consequences arising from its use. It is intended for use only in controlled academic research settings.