This document is very thorough; yet it misses a key point. IMHO, the primary concern here is not the algorithms that produce outcomes and regulating those but what happens to the resulting data streams as well as the raw input for the algorithms. Simply storing huge amounts of video footage means you can later run algorithms as they improve or become available. It's the act of storing and exchanging data that is problematic not running algorithms against that data. You might even capture data before algorithms are invented/available to process that data or use new algorithms against historical archives of data to extract new information.
What are the implications of the right to be forgotten on e.g. security camera footage that is stored for weeks or months or even longer. The raw data contains personally identifiable information, given the right algorithms and access. Mining and correlating information from raw data that so far is considered innocent changes the game completely. Somebody with access to isp data logs, cell tower logs, and security footage in buildings, on streets, etc. can start looking for patterns in that data and cross verify and correlate events between data sets.
Recent history in China teaches us that this is neither science fiction nor an imagined risk. China is quite openly implementing mass surveillance using all means available to them and they are already using it to control their own citizens.
IMHO the focus for legislation should be on legislating not the right to be forgotten, because that is inherently hard to enforce and impractical, but instead on ensuring data capturing parties stick to strict rules with respect to auditing and securing access to that data, retention, transparency of their policies regarding all of this, and making sure serious violations have consequences.
Additionally, any surveillance data used in a court of law needs to have an impeccable auditing record proving it was captured and handled lawfully. One critically important aspect here is also ensuring the data has not been tampered with in any way: algorithms to falsify information are also becoming a risk. Capturing data that may be used for privacy invasive analysis creates a duty to adequately handle and protect that data. This should not be optional.
The key to enforcing this is sousveillance: observing the observers. This should be both legal and common. In a world where there is multi lateral surveillance by multiple groups of people, companies, and governments, hiding undesirable behavior/actions is going to be increasingly hard and you can never be sure that nobody is watching whether you are a citizen, chief of the secret police, or a head of state. Surveillance is a double edged sword and abusing it might be observed by others. The 1984 type scenarios always assume a single evil government doing all the surveillance in secret. So, we should make sure governments themselves are equally exposed to surveillance in order for us to ensure that they do not abuse their level of access to surveillance data.