They can definitely mess with the sequence metadata (date submitted, date published, authors, etc) - but they primarily host NGS sequences which can trivially b

They can definitely mess with the sequence metadata (date submitted, date published, authors, etc) - but they primarily host NGS sequences which can trivially be verified against the original sequencing runs maintained by the researchers, so it's pretty unlikely they've ever tampered with actual genomic data.