Anthropic, OpenAI Leaders Call for Oversight of Self-Improving AI Systems