Jonathan Bha­gan

While there are signs of the AI bub­ble burst­ing, AI has al­ready trans­formed so­ci­ety in un­think­able ways. It is, how­ev­er, im­per­a­tive that T&T be­gin reg­u­lat­ing AI and clos­ing le­gal loop­holes for po­ten­tial­ly dan­ger­ous soft­ware. Nor­mal­ly, an abun­dance of case law from the UK and the Eu­ro­pean Court of Hu­man Rights would pro­vide use­ful guid­ing prin­ci­ples for the field, but since AI af­fects every area of hu­man life, es­pe­cial­ly hu­man rights, it will take many years for enough rel­e­vant case law to be es­tab­lished.

Patent Law has been clar­i­fied by Thaler v Comp­trol­ler-Gen­er­al of Patents, De­signs and Trade­marks, where­by the UK Supreme Court ruled that an in­ven­tor un­der the Patent Act 1977 must be a nat­ur­al per­son and not an AI. The case of Emo­tion­al Per­cep­tion AI Ltd v Comp­trol­ler Gen­er­al of Patents, De­signs and Trade­marks dealt with whether it is pos­si­ble to ob­tain a UK patent for a sys­tem which us­es an ar­ti­fi­cial neur­al net­work (“ANN”) and held that an ANN con­sti­tut­ed a com­put­er pro­gram un­der the Eu­ro­pean Patent Con­ven­tion.

How­ev­er, patent law is but a small frag­ment of what makes up a func­tion­al le­gal sys­tem. The po­ten­tial for hu­man rights in­fringe­ments by new AI tech­nolo­gies is im­mense; how­ev­er, there are on­ly a hand­ful of clar­i­fy­ing cas­es, such as the Eu­ro­pean Court of Hu­man Rights case of Glukhin v Rus­sia, which dealt with fa­cial recog­ni­tion tech­nol­o­gy.

The 1953 Eu­ro­pean Con­ven­tion on Hu­man Rights had an in­flu­ence on the Con­sti­tu­tion of Trinidad and To­ba­go’s Bill of Rights in 1962; hence our courts look to Eu­ro­pean cas­es to in­ter­pret the law, such as in the Tot Lamp­kin case.

In Glukhin v Rus­sia, the Russ­ian state con­vict­ed a peace­ful pro­test­er and used AI-en­hanced fa­cial recog­ni­tion to iden­ti­fy and ar­rest him with­out a clear le­gal frame­work or safe­guards. The court found that this was a breach of the rights to pri­vate and fam­i­ly life cod­i­fied in Ar­ti­cle 8 of the Con­ven­tion, and a breach of the right to free­dom of ex­pres­sion found in Ar­ti­cle 10.

The right to pri­vate and fam­i­ly life found in Ar­ti­cle 8 is the an­tecedent to sec­tion 4 (c) of the Con­sti­tu­tion of Trinidad and To­ba­go, which pro­tects the right of the in­di­vid­ual to re­spect for his pri­vate and fam­i­ly life.

The Eu­ro­pean Court held that the Russ­ian state used AI-en­hanced bio­met­ric soft­ware to process the pro­test­ers’ per­son­al bio­met­ric da­ta in or­der to lo­cate and ar­rest them. The un­jus­ti­fied pro­cess­ing of the ap­pli­cants’ per­son­al bio­met­ric da­ta us­ing high­ly in­tru­sive fa­cial recog­ni­tion tech­nol­o­gy and the im­plic­it sup­pres­sion of the rights of the pro­test­ers by the ar­rest were all con­sid­ered as grounds for the court to find a breach of fun­da­men­tal rights and free­doms.

How­ev­er, the case of Glukhin v Rus­sia has been crit­i­cised due to its fo­cus on pro­ce­dur­al safe­guards, which in ef­fect out­line how to con­duct mass sur­veil­lance with­out breach­ing the law. Some aca­d­e­mics, such as Pro­fes­sor Moni­ka Zal­nieriute, crit­i­cised the Glukhin case as not ground­break­ing, con­clud­ing: “While it is the first de­ci­sion on FRT mass sur­veil­lance, it re­in­forces the court’s long-stand­ing lib­er­al ap­proach that gov­ern­ments may con­tin­ue to de­ploy mass sur­veil­lance regimes where cer­tain (vague) pro­ce­dur­al safe­guards are in­cor­po­rat­ed. The de­ci­sion ac­cepts the idea that FRT mass sur­veil­lance is le­git­i­mate to use in pub­lic spaces in gen­er­al. How­ev­er, Glukhin is al­so em­blem­at­ic of the EC­tHR’s more re­stric­tive ap­proach to more au­thor­i­tar­i­an states, like Rus­sia, and the dif­fi­cul­ty the EC­tHR has in re­in­forc­ing this per­mis­sive ap­proach to mass sur­veil­lance in the face of grow­ing ev­i­dence that the pro­ce­dur­al safe­guards it is ad­vo­cat­ing for are in­suf­fi­cient to pre­vent in­fringe­ment on hu­man rights.”

Giv­en the crit­i­cisms of the di­rec­tion the Eu­ro­pean Court of Hu­man Rights has tak­en to gov­ern ar­ti­fi­cial in­tel­li­gence, com­bined with the lim­it­ed case law avail­able, it is un­like­ly that the court of Trinidad and To­ba­go will have suf­fi­cient guid­ance to deal with its own AI cas­es with­out re­ly­ing on oth­er sources of law. Ar­ti­cle 5 of the Eu­ro­pean Union AI Act has pro­hib­it­ed spe­cif­ic tech­nolo­gies that can re­sult in se­vere hu­man rights breach­es, such as soft­ware that will ma­nip­u­late per­sons via sub­lim­i­nal mes­sag­ing, deep­fake soft­ware and lim­i­ta­tions on AI fa­cial recog­ni­tion. Giv­en the lim­it­ed case law and the lim­i­ta­tions of ex­ist­ing case law, leg­is­la­tion and in­ter­na­tion­al con­ven­tions are the on­ly way to build a re­li­able AI gov­er­nance frame­work.

At­tempt­ing to im­ple­ment a copy of the EU AI Act will not be fea­si­ble, as T&T will first need to adopt a host of oth­er leg­is­la­tion, such as more rigid da­ta pro­tec­tion law first. Giv­en the cost of con­sult­ing fees and train­ing at­tor­neys in AI law, the bur­den of reg­u­lat­ing AI should be shared by our part­ners in the Cari­com or OAS na­tions. An in­ter­na­tion­al con­ven­tion mod­elled af­ter the Coun­cil of Eu­rope Frame­work Con­ven­tion on Ar­ti­fi­cial In­tel­li­gence and Hu­man Rights, Democ­ra­cy and the Rule of Law may be an im­por­tant start­ing point. Fur­ther re­gion­al con­ven­tions on da­ta pro­tec­tion law and oth­er ar­eas of law rel­e­vant to AI may al­so as­sist T&T in mov­ing to­wards a first-world le­gal sys­tem able to fa­cil­i­tate emerg­ing in­dus­tries and busi­ness.