The Bletchley Declaration was signed by 28 countries that agreed on a risk-based approach to frontier AI models, including areas of social protection, health, education, labor. It involved African nations, such as Nigeria, Kenya and Rwanda, countries from the Middle East, including Saudi Arabia and the United Arab Emirates; and major Western economies, such as Canada and the US. Emerging AI policies and frameworks make an attempt to categorize AI systems based on risks, related compliance frameworks and explanations. Such mechanisms are aimed at both regulating and facilitating a human-centered approach to AI systems development, connecting stakeholders and broader society. However, existing approaches to understanding high and unacceptable-risk systems still miss disability-specific vocabulary, scenarios and associated risks, categorization of impairments, spectrums, actions and non-actions, and complex understanding of intersectionality behind it. It includes not only the areas of law enforcement, police, biometrical and public security systems, but less covered areas of silos, misuse or manipulation presented by autonomous systems.