The Midas Project
- Type
- Watchdog or incident database
- Lieu
- États-Unis — United States (EIN 99-2326977)
- Dernière vérification
- 2026-09-22
- Prochaine vérification
- 2027-03-21
Pas encore traduit — affiché en anglais.
Comment les joindre
- Submit a tip — the best-designed whistleblower-adjacent intake in this list
- Signal (encrypted) — usernamemidasproject.10
Ce canal est conservé uniquement comme référence. Sa vérification est absente, date de plus de 180 jours, ou la période de contact n'est pas ouverte actuellement ; ne vous y fiez pas avant une nouvelle vérification.
- General enquiries (explicitly NOT recommended for confidential material)
- Contact form
- Homepage
- Action page
Ce canal est conservé uniquement comme référence. Sa vérification est absente, date de plus de 180 jours, ou la période de contact n'est pas ouverte actuellement ; ne vous y fiez pas avant une nouvelle vérification.
Ce qu'il fait
Watchdog non-profit holding AI companies to their own stated commitments. Projects: Watchtower (tracks changes to and violations of corporate AI safety policies), the Seoul Tracker (whether 16 organisations met the Seoul frontier-safety commitments), The OpenAI Files (documented governance, leadership and culture concerns at OpenAI), and Model Republic (industry influence on policymakers).
Ce qu'ils demandent — dans leurs propres termes
AI companies should keep the safety and transparency commitments they have publicly made, and be held accountable when they quietly walk them back.
Ce qu'ils font
Corporate watchdog work: the Watchtower tracker of silent changes to AI company policies, investigations, scorecards, and the OpenAI Files. Runs modelrepublic.org and openaifiles.org.
Évaluation franche
Tiny, and it absolutely reads its inbox — the single best target in this list for someone with concrete non-public evidence about an AI company's conduct. Note their own advice: if you are an employee considering disclosure, contact the AI Whistleblower Initiative first for independent legal guidance before approaching any journalist or watchdog.
Ce qu'un nouveau venu peut faire
Send a tip if you have inside knowledge of an AI company; otherwise sign up as a volunteer, join the Discord and help with research or amplifying campaigns.
Bon à savoir
Adversarial toward specific named companies. If you work at an AI lab, sending a tip may breach your employment agreement — take legal advice first; the organisation is small and cannot indemnify you.
Préoccupations couvertes
Corporate accountability and commitment-drift — the concern that voluntary safety commitments quietly erode. Agnostic between threat models; it tracks whatever companies promised, which spans misuse, catastrophic risk and governance.
Canal gouvernemental
Indirect but potent — its findings get picked up by press and cited in policy debate; The OpenAI Files fed directly into scrutiny of OpenAI's restructuring.
Conseils
yes — dedicated tip page at themidasproject.com/tips, aimed at industry insiders
Où ils agissent
United States (company-focused, global reach).
Taille
Tiny: founded early 2024 by Tyler Johnston, 'a small core team' plus volunteers. 501(c)(3), EIN 99-2326977.
Dépenses
Not disclosed.