From the VAR Handbook
The aim of the experiment is NOT to achieve 100% accuracy for all decisions as there is no desire to destroy the essential flow and emotions of football which result from the game’s almost non-stop action and the general absence of lengthy stoppages. The philosophy is:
‘minimum interference – maximum benefit’
To ensure that the referee (not the VAR) is the key match official, the referee will ALWAYS make a decision (except a ‘missed’ usually ‘off the ball’ incident), including the decision that no offence has occurred. The referee’s decision can ONLY BE CHANGED if the video review shows a CLEAR ERROR i.e. not ‘was the decision correct?’ but:
‘was the decision clearly wrong?’
Principles
There are twelve principles which are the foundation of the VAR experiment:
1. Video technology will only be used to correct clear errors and for missed serious incidents in defined match-changing decisions: goal, penalty/no penalty, direct red card and mistaken identity (e.g. the referee cautions/sends off the wrong player).
2. The final decision will always be taken by the referee.
3. Video Assistant Referees (VARs) are match officials - any information the VARs
provide to the referee will be treated by the referee in the same way as information
received from an assistant referee, additional assistant referee or the fourth official.
4. The referee must always make a decision regardless of the existence of VARs i.e. the referee is not permitted to give ‘no decision’ and refer the situation to the VAR. If the referee decides not to stop play for an alleged offence, the decision (to allow play to continue) can be reviewed. On rare occasions, when it is unclear whether a penalised cautionable (YC) offence is a sending-off (RC), or who should receive the
sanction, the referee may consult the VAR.
5. The original decision given by the referee will not be changed unless the video
review clearly shows that the decision was clearly wrong.
6. Only the referee can initiate a review; the VAR (and the other match officials) can
only recommend a review to the referee.
7. Whatever the review process, there is no time pressure to review the decision quickly
as accuracy is more important than speed.
8. The players and team officials must not surround the referee or attempt to
influence if a decision is reviewed, the review process or the final decision. A player who uses the ‘review signal’ will be cautioned (YC).
9. The referee should, as far as possible, remain ‘visible’ during the review process to ensure transparency.
10. If play continues after an incident which is then reviewed, any disciplinary action taken/required during the post-incident period is not cancelled, even if the original decision is changed (except a caution/send-off for stopping a promising attack or DOGSO).
11. There is a maximum period before and after an incident that can be reviewed.
12. The VAR protocol, as far as possible, conforms to the principles and philosophy of
the Laws of the Game.
More here
https://www.knvb.nl/downloads/bestand/9844/var-handbook-v8_final