There are no open exploits of exploits that “break through” the protection of virtual machines. Therefore, unless you are working with sensitive information and are not Snowden, you can consider the protection provided by the virtual machine as a whole reliable.
Unless you break it yourself.
First of all, your virtual machine is usually connected to your computer through a virtual network. Thus, a network worm that has infected a virtual machine gets a convenient platform for attacking your main system. If your computer has closed ports, this will not be a problem, but if a router with NAT was selected as the only protection measure, you have problems.
So it's time to deal with a firewall or firewall and find out what permissions you managed to give.
A promising method of attacking a host system is a video card emulator. If you have a choice between several virtual video cards, it makes sense for a virtual machine to limit itself to a standard virtual video card without support for DirectX and OpenGL.
Also a source of danger are virtualization utilities kits for guest operating systems — if you experiment with viruses, they are better not to install them.
In any case, you should not allow a virtual machine unlimited access to the disk.
And finally - it must be remembered that sometimes a virus does not necessarily get out of the virtual machine. Obviously, if you use the same virtual machine to experiment with viruses and to pay for anything from the card, the virus will drag out your billing information in the same way as he would on a real computer.
A less obvious scenario is that a virus that has infected a virtual user may include it on the botnet and DDOS another site from it. Or mine cryptocurrency for your electricity.
Update Intel processors found a Meltdown vulnerability, which, in the absence of patches closing it, allows any process to arbitrarily read any places in RAM. Therefore, it is better not to work with important data when running a virtual machine with viruses if you have Intel.