Current AI, a nonprofit founded in February 2025 by Martin Tisne and recently led by CEO Ayah Bdeir, is advancing a mission to construct a public, open-source alternative to proprietary artificial intelligence systems. Backed by $400 million in committed funding, including a $100 million seed from the French government alongside contributions from the Ford and MacArthur foundations, DeepMind, and Salesforce, the organization aims to democratize access to AI while prioritizing linguistic diversity and data sovereignty. Bdeir, who previously directed AI strategy at Mozilla, emphasizes that while major tech firms develop multilingual models primarily for market expansion, Current AI focuses on preserving endangered languages and empowering underrepresented communities. The initiative has rapidly transitioned from concept to deployment. At the February India AI Summit, Current AI partnered with the Indian government’s Bhashini division to release Suno Sutra, an open-sourced, offline-capable device supporting 22 Indian languages. Designed to function without internet connectivity, the hardware enables local users to access AI tools regardless of language barriers. Simultaneously, the nonprofit addressed global linguistic gaps through a $3.2 million grant round distributed to four regional organizations. Projects include Masakhane’s development of multilingual AI datasets for Kenyan health and agricultural sectors, the Lebanese Institute for Worldmaking’s community-controlled digitization of Arab cultural heritage, Brazil’s Portal sem Porteiras initiative for offline Indigenous Amazonian tools, and the African Internet Rights Alliance’s cross-continental AI accountability audits. In Geneva, at the AI for Good Summit, Current AI unveiled Alpha Chat, an open-source conversational interface developed over seven weeks by a consortium of ten entities, including Hugging Face, Mozilla, and the MIT Media Lab. The platform modularizes core infrastructure, allowing independent contributions for language models, safety protocols, and computational resources. To further expand its stack, Current AI recently partnered with Tokyo-based Sakana AI to develop a shared open-source framework tailored for Japanese linguistic and cultural contexts, with broader applications for Global South communities. Central to Current AI’s operational model is strict data localization and community consent. Bdeir rejects centralized Silicon Valley data extraction, advocating instead for on-device storage, localized model training, and participatory design pipelines that allow communities to halt or modify data usage at any stage. Rather than pursuing rapid global scaling, the organization prioritizes targeted, culturally resonant implementations. The underlying philosophy mirrors the open architecture of the early web, ensuring that technological advancement remains publicly accessible, interoperable, and governed by the communities it serves rather than proprietary corporate interests.