MiniMax Group Inc. Prospectus (Global Offering)

Key Findings

• — Risk factors acknowledge that 'certain AI companies have been sued in connection with allegations that their chatbot products contributed to users' self-harm or suicide' as an industry risk applicable to the company's own AI-native products
• — Describes a staged AI safety and alignment methodology: data curation (annotation, filtering, and detoxification of training data), model training (supervised fine-tuning and reinforcement learning to reinforce refusal behaviors), and deployment (runtime classifiers scanning for harmful intent, including self-harm-related content, with escalation to a safety team)
• — States the company performs 'graded risk assessments, particularly for high-risk areas such as mental health or self-harm' at the model design stage, with continuous monitoring and escalation of high-risk interactions during operation
• — Discloses that Talkie and Xingye include a 'Teen Mode' enforcing stricter controls for underage users, including a 10pm-6am access curfew and disabled AI-character creation/search/editing features
• — Discloses that a prior version of the Talkie app was temporarily removed from Apple's App Store in certain jurisdictions from mid-December 2024 to mid-February 2025 (Apple did not specify a reason; the company states this was not due to product default or illegality); describes content-moderation and risk-management adjustments made during the removal period, including enhanced automated-plus-human content review, before the app was reinstated

Methodology Notes

A securities-law-mandated IPO prospectus filed with the Hong Kong Stock Exchange (filed on or around 2025-12-31; the company listed 2026-01-09). Safety-governance content is drawn from the Risk Factors and Business sections of a 716-page filing rather than a dedicated technical safety report, and is self-disclosed under regulatory disclosure obligations rather than published as an independent research or transparency document.

Sources