It would appear that the current policy at GenBank is to represent all genomic sequences as DNA, even though this is not made explicit in any of the easily retrievable documentation on their website or in their publications. The way in which one determines the nature of the genome is from the first (LOCUS) line in the report, which, in the case of SARS-CoV-2 is:
LOCUS NC_045512 29903 bp ss-RNA linear VRL 30-MAR-2020
The ss-RNA indicates a single-stranded RNA genome.
The only indication I could find that this might be GenBank policy was a reference to the fact that a separate RNA division of GenBank used to exist, but was discontinued in 1999:
The RNA division of GenBank was removed in release 113.0 (August 1999). Sequences that were previously in the RNA division have been moved to the appropriate organismal division.
The genomes of certain RNA phages were originally sequenced as RNA (as were bacterial rRNAs and 18S rRNA) — that of phage MS2 was the first in 1976. The GenBank entries for these are currently in DNA format. Thus V00642.1 contains references to the 1976 MS2 paper by Fiers et al., but is in DNA format. As GenBank was not established until 1982 it is not clear whether the original entry in GenBank was in RNA or DNA format.
I wrote to GenBank to ask them whether there was a specific policy and if so when it had been introduced. I had a reply from someone on their help desk to the effect that GenBank policy was that of International Nucleotide Sequence Database Collaboration, the shared rules of which includes table "7.4.1 Nucleotide base codes (IUPAC)". This states that "t" can represent thymine in DNA and uracil in RNA. However my respondent could not find out whether this representation was used in GenBank for RNA viruses from the beginning or introduced later. As originally GenBank releases were on CD-ROM this is difficult to check.