Forg3t Protocol

We built Forg3t after realizing something uncomfortable: AI systems are great at remembering, but terrible at proving they forgot something. Today, most “deletion” workflows rely on internal claims, retraining promises, or policy statements. That is not enough when regulators, courts, or users ask for evidence. Forg3t treats unlearning as something that must be proven at the behavior level. We run targeted unlearning, verify suppression with adversarial prompts, and generate cryptographic proofs that can be independently validated. This launch is our first step toward making AI forgetting auditable, not trust based. We would love feedback from people building AI systems under real regulatory pressure.