Fireside Chat: Jeff Moss and Ruimin He

AI requires human oversight and “humans in the loop” - machines alone cannot be fully trusted or held accountable

Cultural context and societal values must be reflected in AI systems and their governance - different regions/countries will approach AI regulation differently based on their values

There’s an ongoing “land grab” in AI with companies racing to accumulate data and compute resources, creating concerns about centralization of power

Current AI systems are better at tasks with clear feedback loops (like offense/attacks) compared to defensive applications where success signals are less obvious

Identity and authentication remain critical challenges - AI systems need better ways to verify human users while balancing security and accessibility

Code security and memory safety are key areas where AI could help, but still requires human oversight and testing

Regulations need to be sector-specific rather than broad horizontal rules, as risks and benefits vary significantly by use case

Small countries like Singapore see AI as an opportunity multiplier to overcome resource limitations

Model training data needs to reflect local cultural context and values to be effective

Building public trust and adaptability is critical - populations need to be prepared for AI-driven changes while maintaining human agency

Global cooperation and standards are needed for AI governance, but individual countries must also experiment with different approaches

Current anti-bot/CAPTCHA systems may be counterproductive as they’re becoming harder for humans while being easily defeated by AI