Alan Chan suggests that AI agents should identify their inference provider to help trace incidents back to their source