A new coalition called the AI and Multimedia Authenticity Standards Collaboration (AMAS) has been formed, led by the IEC, ISO, and ITU, to develop standards that enhance trustworthiness in AI and digital media.

AMAS is working on future standards that include digital watermarking, content trust, and multimedia authentication to combat issues like misinformation, deepfakes, and bias.

The coalition aims to address misuse of AI-generated content and promote user trust across the digital ecosystem, with standards addressing content provenance, authenticity, and human rights.

Gilles Thonet from IEC emphasizes the importance of defining what constitutes an AI system, highlighting the complexity of interactions within AI technologies.

The standardization process is evolving to include non-technical professionals such as ethicists and social scientists, alongside engineers, to better address societal impacts.

Rapid AI advancements, like agentic AI, challenge standards bodies to develop adaptable and timely responses to keep standards relevant.

The initiative was announced at the 'AI for Good' Global Summit in Geneva, with leadership from IEC, ISO, and ITU, signaling a global effort to establish trustworthy AI standards.

Metzger underscores the need for formal assessments, including audits and third-party verification, to ensure compliance with emerging standards.

While current standards are voluntary, market pressures are encouraging companies to adopt them to boost credibility and ecosystem integration.

Despite AI's rapid evolution, Metzger remains optimistic about society’s ability to implement governance systems that adapt and mitigate potential harms.

The coalition’s efforts include addressing challenges like deepfakes, bias, and misinformation through standards focused on content provenance, trust, and authenticity.

Standards under development include guidelines for content provenance, trust, authenticity, asset identifiers, and human rights declarations, with foundational standards already published in 2020.

Recent standards include JPEG Trust Part 1, ensuring trust and authenticity in JPEG images, and Content Credentials for traceability and verification of digital content.