The Midas Project

The Midas Project is an AI safety advocacy nonprofit founded in early 2024 by Tyler Johnston. The organization's name references the King Midas myth, drawing a parallel used by AI safety researchers like Nick Bostrom and Stuart Russell to illustrate the dangers of AI systems that pursue goals with unintended consequences. The nonprofit describes itself as a social movement demanding transparent, ethical, and safe artificial intelligence that benefits everyone. Johnston came to AI safety from animal welfare advocacy, where he worked as a corporate communications specialist at The Humane League and as a corporate engagement research fellow at the Good Food Institute. He holds a Bachelor's degree from Harvard College. Before founding The Midas Project, he received a $35,000 grant from Open Philanthropy to support his career transition into AI policy. He brought the tactic of targeted corporate campaigns, which had successfully pushed food companies to adopt cage-free egg practices, to the AI safety space. The organization launched in mid-2024 with less than $25,000 in funding and volunteer support. It received a $31,000 grant from the Survival and Flourishing Fund in 2024 along with a prior $10,000 SFF Speculation Grant. Johnston has been the only full-time employee, earning approximately $25,000 per year, with the majority of campaign work carried out by unpaid volunteers. The Midas Project's flagship initiative is the AI Safety Watchtower, launched in mid-2024, which tracks 16 companies including OpenAI, Google, Anthropic, Meta, and xAI, monitoring hundreds of policy documents and web pages for changes. When evidence shows backtracking or inadequate safety controls, the organization documents these gaps and publicly presses for corrective action. In February 2025, it launched the Seoul Commitment Tracker to monitor whether 16 AI companies were fulfilling safety commitments made at the 2024 AI Safety Summit in Seoul, finding that no company earned better than a B-minus grade and six received failing marks. The organization is best known for co-publishing The OpenAI Files in June 2025, a 50-plus-page investigative report produced in collaboration with the Tech Oversight Project. The report documented concerns about OpenAI's governance practices, leadership integrity, and organizational culture based on hundreds of documents, employee testimonies, and media reports. The publication drew significant media attention and led to OpenAI issuing subpoenas to The Midas Project and Johnston personally as part of ongoing litigation with Elon Musk. In January 2026, The Midas Project launched Model Republic, an online publication investigating AI industry lobbying and its influence on public policy, with its inaugural report examining Andreessen Horowitz's policy activities and investment record. The organization has been featured in Fortune, Fast Company, TechCrunch, Yahoo Finance, and other major outlets, and Johnston has appeared on the Future of Life Institute podcast discussing OpenAI's treatment of its critics.