For True Audio-to-Audio Artificial Intelligence: Do Not Leave 700 Million Oral-Dependent People Behind

1. Orality and Digital Exclusion: Orality remains the most direct and universal form of communication on the planet. Most people naturally rely on the voice to exchange ideas, gather information, and learn because it is immediate, intuitive, and profoundly more inclusive.
Yet, more than 700 million audio-literate individuals worldwide have no choice but to rely on orality. Today’s audio-to-audio machine intelligence systems exclude these 700 million audioliterate people, widening the global digital and linguistic divide.

2. Why This Matters: An audio-to-audio machine intelligence capable of generating, interpreting, and transmitting voices and signs in all languages must:

Respect international law and human rights, including Articles 19 and 26 of the Universal Declaration of Human Rights, as well as the UN Guiding Principles.

Protect personal data, in strict compliance with GDPR and Council of Europe Convention No. 108.

Ensure full transparency of training data and algorithms, with clear civil and criminal accountability in case of harm.

Preserve and promote linguistic diversity, while forbidding any use for abusive surveillance or disinformation.

Publish all datasets and code. Sharing only weights/biases or code while concealing data is not real transparency.

3. Our Demands for Responsible AI: We call for public audio machine intelligence that:
Records all data on a public, immutable blockchain, with clearly defined multi-stakeholder governance enabling autonomous, independent audits.
Identifies machine mirages, their limits and biases, and establishes multi-level, multi-scale responsibilities for all actors.

Embeds multi-scale ethics from design through the full lifecycle, with robust mechanisms for auditing, governance, and sanction to correct failures.
Provides universal access, including in the most remote areas, with hardware adapted for everyone.

Is funded through international public funds, voluntary contributions from member states, multilateral partnerships, and transparent collective financing mechanisms, ensuring free and equitable access for all.

It must  embrace environmental responsibility through consuming less, prioritizing renewable-energy-powered data centers, adopting sustainable computing practices, and ensuring sustainable AI from design to post-deployment across its entire lifecycle and worldwide, so that technological inclusion goes hand in hand with ecological and social co-construction &  co-development.

Establishes, under the United Nations, an international regulatory framework guaranteeing safety, ethical standards, decision traceability, and privacy protection, so this technology serves the common good, strengthens justice, education, and universal communication.

4. Call to Action: We urge individuals, governments, international institutions, and companies across all sectors to support this petition. Act now so that everyone, regardless of language or ability, can benefit from inclusive, ethical, and responsible audio-and-sign machine intelligence.

Resources:
1. Universal Declaration of Human Rights -  Article 19
2. Universal Declaration of Human Rights -  Article 26
3. United Nations Guiding Principles on Business and Human Rights
4. General Data Protection Regulation (GDPR)
5. Council of Europe Convention No. 108

6. Breaking the Barriers of  "Text-hungry” and "audio-deficient” AI.
This article presents an audio-to-audio, text-free machine intelligence framework, designed to serve audio-literate populations, particularly in Africa.
📄 https://arxiv.org/abs/2506.02443

7. Machine Mirages:   a proposal for a systematic evaluation of these failures.
📄 https://arxiv.org/abs/2506.13990