18 Countries, Including the US and UK, Sign Agreement to Ensure Safe AI Design | WPN World People News

The U.S., U.K., and 18 other countries announced an international agreement on artificial intelligence (AI) safety on the 26th to ensure that AI is protected from malicious use and to promote the development of "design-safe" AI systems by corporations to ensure the safety of their customers and the general public.

Eighteen countries have agreed that companies designing and using AI should develop and deploy it in a way that ensures customers and the wider public can use it safely and without abuse. The agreement is the first of its kind on a global scale. The agreement addresses four key areas of the AI system development lifecycle: secure design, secure development, secure deployment, and secure operations and maintenance, each of which highlights considerations and mitigations that can help reduce cybersecurity risks in an organization's AI system development process.

The agreement is non-binding and contains mostly general recommendations, such as monitoring AI systems for misuse, safeguarding data from tampering, and vetting software vendors. The framework tackles the issue of how to prevent AI technologies from being manipulated by hackers and suggests that organizations should only release models that have passed security tests. But the guidelines don't address several thorny issues, including how AI should be used appropriately or how data from input models should be obtained.