A coalition of global policymakers and technology developers signed a safety pact in France aimed at preventing the potential loss of control over frontier artificial intelligence. Under the resulting “Paris Call for AI Safety and Security,” 10 companies, including OpenAI, Microsoft and Anthropic, along with six governments, pledged to implement fail-safe mechanisms that allow for the shutdown of AI models during their training and internal use phases.
The agreement was the primary outcome of a summit convened by President Emmanuel Macron that brought together about 50 heads of state and AI executives to discuss the risks of advanced models. While the pact establishes guidelines for incident reporting and public risk evaluation, some participants, including researcher Yoshua Bengio, argued the framework lacks specific deadlines and required mandatory independent audits to prove model safety before deployment.
