On February 2, the EU AI Act, which came into force in August, reached the first compliance deadline. These are provisions that contain rules that allow the bloc to legalize biometric mass surveillance.
This is happening via Article 5, that on the face of it seeks to introduce protections against using AI systems that represent “unacceptable risk.” But, there are also exceptions to this, defined in the act as the “strictly necessary” cases when those same systems are allowed to be used.
It is this that gave rise to the fear that one of the consequences of the AI Act is to in fact legalize some highly contentious deployments of biometric data-fueled mass surveillance.
Article 5 prohibits real-time remote biometric ID systems from being used by law enforcement in public spaces – but only “unless and in so far as such use is strictly necessary.”
The “strictly necessary” instances are described as those when law enforcement is searching for “specific victims” of crimes like abduction, trafficking, and sexual exploitation, but also when they look for missing persons.
The second definition gets less precise in scope as it allows for AI surveillance systems to be used to prevent crime. This includes a threat to life or physical safety of individuals that is deemed to be “specific, substantial, and imminent” – or threat of a terrorist attack that law enforcement decides is “genuine and present” but also – “genuine and foreseeable.”
Lastly, the AI Act treats as “strictly necessary” to exempt from prohibited AI activities the following: “Localization or identification of a person suspected of having committed a criminal offense, for the purpose of conducting a criminal investigation or prosecution or executing a criminal penalty for offenses.”
These are referred to in Annex II and are punishable in EU member countries for a maximum period of at least four years.
In other words, quite a lot of “strictly necessary” scenarios and interpretations that allow for mass biometric surveillance can be derived from the rules.
To facilitate enforcing the act, the EU came up with its “AI Pact” – another of the bloc’s “voluntary pledges” signed by over 100 companies, Amazon, Google, and OpenAI among them – but not Meta and Apple.
Some observers note that companies critical of the law expect clarity regarding standards and guidelines.
