**In 200 words or less, describe an under-discussed (i.e. at the very least, it should not already be implemented) AI governance proposal you'd be excited to see governments or frontier companies implement, and why.** There are mostly 2 governance systems that come to my mind, one is simple but less perfect and the other one is complex but close-to-perfect. I'm going to describe quite-complex but close-to-perfect. Ideally, the governance system should have 3 components: 1. Justice - offering rewards and penalties for good/bad actions and/or complying / not complying with agreed rules. 2. Equality - some form of partial wealth redistribution to avoid concentration of power. 3. Truth - deciding what is true for the purpose of achieving justice and equality. Truth should be achieved with a prediction market. Using that prediction market, people agree on what rules are best for the collective. They also contribute knowledge about the instances of compliance/violation of the rules. The people's incentive to contribute knowledge to that prediction market is that they make money (or "reputation", "social score" or something else) when they are right. The prediction market is decentralized, i.e. the database is distributed among the computers owned by the participants. Otherwise, whoever is in charge of that prediction market can cheat. Justice and equality is achieved through indirect reciprocity. More information: https://theoreticalexplorer.com/Alignment+between+humans/JET/JET+(a+decentralized+governance+system)