IT Home announced on February 3 that the first compliance dead Whatsapp for the European Union ' s Arti ficial Intelligence Act (AI Act) has come into effect on February 2 local time .
This means that EU regula tors can now ban AI systems that they consider to be " una c cep table risks " or harm ful .
The bill offici ally came into force on August 1 st , a compre hen sive framework for artificial intelligence reg ulation that the EU had approved in March last year after years of ena ct ment .
The first compliance requir ement of the bill , which came into force this time , main ly address es the " una c cep table risk " AI applications specified in Article 5 of the bill .
These applications include AI used for social sc oring , AI used to manipula te individual decisions in sub cons cious or fraud ulent ways , AI used to explo it weak nesses such as age , disa bility or socio econ omic status , AI used to predic t crime from appe arance , AI used to estim ate individual character istics (su ch as sexual orient ation) using bi omet ric technology , and AI used to estim ate individual character istics (su ch as sexual orient ation). These include AI that colle cts " real - time " bi omet ric data in public places and uses it for law enfor cement ; AI that tri es to estim ate people ' s emoti ons at work or school ; and AI that cre ates or ext ends fa cial recogn ition databas es by incorpor ating network or security camera images .
Any company , regard less of where it is head quar tered , using any of the above AI applications within the EU could face a fine of up to € 35 million or up to 7% of its annual revenue for the previous fiscal year .
The higher standard is used .
Rob Sum mer , head of technology at UK law firm Sl ug ger & May , point ed out in an interview that the fine would not take effect immediately .
Organiz ations must be fully comp liant by February 2, but the next critical period is August .
At that time , we will know who the competent authorities are and the fines and enfor cement provisions will also come into force . "
In September last year , more than 100 companies signed the EU Arti ficial Intelligence Treaty , volunt arily commit ting to start apply ing the principles before the AI bill comes into force .
The signat ories are committed to recogni zing that they include Amazon , Google and Open AI , which could be classi fied as a high - risk AI system .
However , some science and technology gi ants , including Meta and Apple , have not signed the deal , nor has the French artificial intelligence vent ure M istr al .
Sum roy says that given the nature of the use cases listed , most companies are not enga ged in any of these practices .
He stress ed that the main concern of each organization was whether clear gu ideWhatsapps , standards and standards of conduct could emer ge in a tim ely manner and whether they could provide clar ity on the compliance side to the organization .
The bill provides for excep tions to some prohib itions .
For example , law enfor cement will allow systems to collect bi omet ric data in public places to help search for ab du ction victims and prevent " spe cific , serious and im min ent " thre ats to life .
However , such exemp tions need to be author ized by the correspond ing regula tory body , and the bill stress ed that law enfor cement ag encies cannot make decisions that " have an ad verse legal impact " based sole ly on the output of these systems .
The bill also provides for excep tions for systems that estim ate emoti ons in work pla ces and schools , such as systems used for treatment , if there are " med ical or safety " reasons .
The Commission has previous ly said it would publish more gu ideWhatsapps " in early 2025 " after consult ing stak ehol ders in November , but these gu ideWhatsapps have not been made public .
Sum roy said he still doesn ' t know how other current laws inter act with the AI bill ' s prohib itions or related provisions .
Later this year , the situation may become clear er as the law enfor cement window approach es .
Sum roy emph asi zed that organizations must remember that the over sight of artificial intelligence is not in isola tion .
Other legal frame works , such as GDPR , N IS 2, and D ORA , would inter act with artificial intelligence legislation and pose potential challenges , particularly in terms of over la pping event notification requirements .
Under standing how these laws are linked together is as important as understanding the AI bill itself .
Release date: 2024-01-04 20:08:01
