A look at OpenAI's “Democratic Inputs to AI” grant program, which awarded 10 teams $100K each to design ideas and tools to collectively govern AI systems
Billy Perrigo / TIME : X: @time , @lopezunwired , and @billyperrigo LinkedIn: Teddy Lee and Laura MacCleery X: @time : The company wants to align its AI to “human values.” Is that even possible? https://time.com/... Paul Lopez / @lopezunwired : Colin Megill, CEO of Polis, announced in May 2023 the collaboration with OpenAI to integrate their AI-powered public deliberation platform, leading to the “Democratic Inputs to AI” initiative. https://time.com/... Billy Perrigo / @billyperrigo : OpenAI wants to align its AI to ‘human values.’ But whose? My latest feature tells the inside story of OpenAI's ongoing effort to make AI more ‘democratic’ https://time.com/... LinkedIn: Teddy Lee : 🌍 TIME published a nuanced article about the work that my team at OpenAI is doing to ensure that our models reflect the diverse values … Laura MacCleery : A new article in Time offers the latest take on Open AI's overly publicized experiments with “democratizing” AI. …
Context & Ripple Effects
OpenAI first solicited proof-of-concepts for a democratic process to set AI rules through its $100,000 governance-grant program. The awards turn that call into a portfolio of 10 concrete experiments in collective AI governance.
The program now sits alongside OpenAI’s Collective Alignment team, which is intended to gather and use public input in shaping model behavior. Polis’s involvement ties the effort to a public-deliberation platform rather than a purely internal research exercise.
First-order effects
- Ten teams receive $100,000 each to develop governance ideas and tools, giving OpenAI a set of externally developed approaches to evaluate.
- OpenAI gains a formal channel for testing how collective input, including work connected to Polis, could inform its stated alignment goals.
Second-order effects
- The immediate challenge shifts from collecting proposals to deciding which forms of public input can be translated into model rules and operating processes.
- The grants make participatory governance a more tangible part of OpenAI’s alignment agenda, raising the bar for its internal teams to show how outside input is used rather than merely solicited.
Third-order effects
- If such pilots are incorporated into model-development processes, AI governance could increasingly be treated as an operational product function—alongside training, safety testing, and deployment controls.
- The effort also illustrates a broader legitimacy test for frontier-model developers: public participation may become a durable complement to company-led alignment decisions, though its influence will depend on who participates and how outcomes are applied.
The trend: AI labs are moving from abstract alignment commitments toward repeatable mechanisms for incorporating public input into model governance.