It has become apparent that the use of real-time genetic data generated by AI without official authorization will be prohibited for law enforcement and national security bodies in the European Union.
After three weeks of negotiations, the German parliament and EU member states agreed on Friday to a conventional directive. The final document will remain undisclosed until a “cleaning up process” is completed, with operational details only made public on Monday.
Except in cases of specific severe crimes, terrorist threats, or vital searches for individuals, extensive surveillance akin to “big brother” will be limited in public and private areas, including recreational spaces and sports facilities. Even in such situations, approval from a magistrate or another impartial figure is required before police action.
In exceptional circumstances, like imminent terrorist threats, law enforcement cannot utilize AI biometric technologies without judicial consent. However, an EU entity knowledgeable about the subject mentioned the possibility of obtaining approval within 24 hours and conducting a “prior basic rights impact assessment” overseen by a relevant expert.
Additionally, authorities must inform data protection and market oversight bodies. If authorization is denied, the AI tool must be immediately deactivated, and any data on suspects must be erased.
Authorities stress that these measures aim to prevent “predictive policing,” a practice that MEPs feared could lead to discrimination when combined with cultural monitoring. According to an EU source, this rule prohibits targeting individuals as suspects solely based on an algorithm’s classification.
A mutual agreement has been reached between the Union and MEPs on a comprehensive list of 16 major crime categories where security measures may be applied.
These exceptions include violence, murder, sexual assault, organized or armed robbery, serious physical harm, child exploitation, kidnapping, hostage-taking, crimes falling under the jurisdiction of the international criminal court, hijacking of aircraft or vessels, human trafficking, organ trafficking, drug trafficking, weapons, nuclear materials, and involvement in criminal organizations linked to these offenses.
Each country must comply with these AI limitations within six weeks of the laws being promulgated in the International statute books or the Official Journal.
Further restrictions will be imposed in the new EU regulations governing artificial intelligence to protect society and individuals from the dangers associated with AI.
These restrictions encompass AI systems or applications that “influence human behavior to undermine individuals’ autonomy,” such as voice-activated toys that encourage risky behavior in children.
Moreover, there will be a ban on AI platforms that allow governments or companies to implement a “social credit” system, similar to China’s social credit system.
Lastly, the EU emphasized these measures due to concerns that AI-derived data could be used to exploit individuals based on their age, disabilities, or social and economic vulnerabilities. This could result in situations where certain individuals are denied access to establishments due to their “social score” or status, or where employees are dismissed based on similar criteria.
In work settings, the use of AI for personal identification, involving real-time analysis of individuals’ facial expressions to assess stress or fatigue, will also be prohibited.
MEPs supporting these bans were steadfast in their efforts to prevent the EU from becoming a surveillance state resembling China, where, for example, traffic police can intervene if they suspect a driver is fatigued.
