> where does the bias come from?
Don't put any grid-leak on a tube. For small types with significant external plate resistor, the grid will tend to drift a part-volt negative and the operating conditions suitable for small-signal work. With infinite resistance it is not real stable, and a strong transient can throw it so far off that it quits working. Values like 1Meg are uttlerly stable and essentially zero bias. 10Meg is a workable "contact bias" for most of the small tubes. 6.8Meg may have been on sale that week.
While there are contact-potential effects, this leak-bias is more about the spread of initial energies of the cathode electrons. Some will jump into the grid even when it is a little negative and make it more negative. Too negative, they stop going there. In healthy tubes you get a few tenths volt negative bias.
This is OK for dynamic mikes and early e-guitars. As hotter sources came out, most G-amps went to cathode bias. This amp's divider input also takes somewhat higher inputs, at the cost of S/N. If you care, buy something else.