I think his comment is exaggerated.
There is no energy violation, and the energy is conserved in all vertices individually.
The virtual particle however is an off-shell particle, which means that its 4-momentum squared is not equal to the particle's rest mass. In formula what is violated is the equality : [itex]E^2 - |p|^2 c^2 = m^2 c^4[/itex].
The mass would have to be replaced by [itex]s= q_w^\mu q_{w \mu} \ne m_W^2[/itex].
If someone wants to use the Uncertainty principle to describle virtual particles he generally says that there can be fluctuations in energy [itex]\Delta E[/itex] within some time [itex]\Delta t[/itex], that obeys the [itex]\Delta E \Delta t \sim \hbar[/itex]. This can help you get an insight about how the particle pops up into being. Of course it's not a real particle, meaning that you cannot observe it. In order to observe it you have to use higher energies and make it pop up into being real (that's a resonance)