> A data file with an invalid format can cause a driver to crash badly enough to take out the entire OS?
The use of "driver" is - confusing. Unlike, say, a graphics card driver, antimalware has deliberate code that will poke itself into everywhere - if it goes truly apeshit it could amazing amounts of damage! So, yes, it probably could crash the OS.
Although it is not crashing the OS, it is preventing the OS booting in the first place. Not a subtle difference, as it is almost certainly inserting itself as early into the boot sequence as it can, so probably not that much of the rest of the OS will be operational. Yes, that puts it into a very privileged position (as it would need to be if it can do what it claims).
That said, the invalid file format is totally inexcusable. Parsers should be robust enough by now to just report back "nope".[1] And at this level you do absolutely everything to validate before you actually load and use a "channel", whatever that really is.
Someone probably said "we can not parse and validate, that will slow down the boot; it was checksummed in the download and we can totally trust the copy on the server to have been validated"!
[1] although I have seen morons who believe that a parse error should raise an exception and then you get into all the crap about not catching the exception or letting it pass back up because it is really that layer's responsibility/interest... How the ¢π$¥{¶ is a parse error ever an exceptional situation? Then again, there are lookup tables out there where a "not found" raises an exception! FFS.
Biting the hand that feeds IT
