You can have access to any kind of dataset through Google's dataset search engine. It has got a compilation of more than 25 million datasets. I think you'll find your required dataset as well.
This dataset includes CSV files which contain the tweet IDs. The tweets have been collected by the LSTM model deployed here at sentiment.live. The model monitors the real-time Twitter feed for corona virus-related tweets. As per the Twitter Developer Policy, it is not possible for me to provide information other than the Tweet IDs (this dataset has been completely re-designed on March 20, 2020, to comply with data sharing policies set by Twitter). Note: This dataset should be solely used for non-commercial research purpose.
We have compiled a large and new dataset of COVID-19 CT scans of patients. You can use it at the following link:https://www.researchgate.net/publication/341804692_A_Fully_Automated_Deep_Learning-based_Network_For_Detecting_COVID-from_a_New_And_Large_Lung_CT_Scan_Dataset/stats
In the "Coronavirus Disease Research Community - COVID-19" you may find lots of datasets, or even deposit your own: https://zenodo.org/communities/covid-19/
So far, all the available data about COVID 19 are non-deterministic; the error ratio of any given statistical information is significant. The main reasons for those errors are simply confined as follow:
1. Some tests, which have been done to some people, were considered positive in spite of the detected virus was the classical Corona; not the new one.
2. Some people are really infected (maybe me or you!) but the symptoms don't appear on these people. Thus, they are not included in the total number of infected people.
As the number of people from the two categories is really huge, how to utilize the available statistical information to make a reliable study that predicts the evolution of this epidemic in the future? The results of such a study should contain a vast deviation between the predicted and the real state in the future.
Someone may say, we may deal with this stochastic data by using some probabilistic techniques. But, as we are unable to model the process error or its associated observation error, how to build a dynamic system describing the predicted state?
In any case, at the end of this scaring period, we will discover that a major part of these available studies is unrealistic. Because the real evolution curve of COVID 19 will show us its real face.
You can find an updated list of Coronavirus (Covid-19) dataset in the following link: https://www.researchgate.net/post/updated_list_Last_updated_July_10th_2020_of_Coronavirus_Covid-19_dataset_and_other_Research_Resources
Moreover, Here is an open dataset of 100 segmented axial CT slices from ~60 Italian patients:
It probably depends on what type of dataset you are searching for. For deep learning and image classification or similar tasks, you can use the dataset from the following GitHub link - https://github.com/lindawangg/COVID-Net/blob/master/docs/COVIDx.md. They collected the COVID-19 chest X-ray images from different sources and gave a guideline on how you would use it.