Anthropic, an AI safety company, has called for AI laboratories to establish a coordinated plan to pause development if significant risks from artificial intelligence emerge.