Geisha Kow­lessar-Alon­zo

Se­nior Re­porter

geisha.kow­lessar@guardian.co.tt

The Cen­tral Bank of Trinidad and To­ba­go is an ear­ly adopter of Ar­ti­fi­cial In­tel­li­gence (AI), but the reg­u­la­tor of fi­nan­cial in­sti­tu­tions re­mains cau­tious about its use.

Keisha Lash­ley, as­sis­tant man­ag­er of In­for­ma­tion and Cy­ber Se­cu­ri­ty at the Cen­tral Bank, made this point while de­liv­er­ing re­marks at a we­bi­nar and pan­el dis­cus­sion ti­tled, “Ar­ti­fi­cial In­tel­li­gence, Cy­ber Re­silience and Cy­ber Se­cu­ri­ty” host­ed by the bank yes­ter­day. The we­bi­nar was held in col­lab­o­ra­tion with the Bank for In­ter­na­tion­al Set­tle­ments (BIS).

Re­gard­ing the im­ple­men­ta­tion of AI tech­nol­o­gy, Lash­ley said at the Cen­tral Bank, like many oth­er in­sti­tu­tions, many em­ploy­ees are very in­ter­est­ed in util­is­ing AI.

“Cer­tain de­part­ments are us­ing it and the process we fol­low is that if a busi­ness user is in­ter­est­ed in us­ing any AI-re­lat­ed tech­nol­o­gy, as a mat­ter of fact, there’s a process that they must ad­here to where the se­cu­ri­ty team would do a re­view and we would give feed­back on it.

“We use it in the IT de­part­ment. A lot of new tech­nol­o­gy is re­lat­ed to ma­chine learn­ing, so that is al­ready ful­ly en­trenched in some of our tools. So, I’d say we are ear­ly adopters. We are a cen­tral bank and we tend to be a bit more cau­tious than oth­er in­sti­tu­tions. There’s in­ter­est and we try to bal­ance that in­ter­est with do­ing it se­cure­ly,” Lash­ley added.

Dur­ing the dis­cus­sions, the is­sue of the “threat land­scape” with­in T&T was raised as it re­lates to cy­ber­se­cu­ri­ty.

Lash­ley said ran­somware is get­ting “a lot of at­ten­tion.”

“What we see hap­pen­ing lo­cal­ly is out­side of com­pro­mis­ing your en­vi­ron­ment. They (cy­ber­at­tack­ers) are ex­fil­trat­ing da­ta and try­ing to put that ad­di­tion­al pres­sure on the in­sti­tu­tions. So, they are al­ways seek­ing ways to mon­e­tise and that’s what we see hap­pen­ing,” Lash­ley added, as she ad­vised fi­nan­cial in­sti­tu­tions that they need to keep their guard up at all times.

“They (cy­ber­at­tack­ers) of­fer mal­ware as a ser­vice, mul­ti­func­tion­al mal­ware. So it is tough as de­fend­ers, but we have to try and keep pace. We have to evolve and as­sess and keep grow­ing even though the at­tack­ers do their bit,” Lash­ley said, not­ing that T&T is not unique when it comes to cy­ber­crimes.

She said at­tack­ers of­ten look for “soft tar­gets” whether it is a cred­it union, an in­sur­ance com­pa­ny or a bank.

While stat­ing that she did not have any da­ta to point to any in­sti­tu­tion be­ing more at risk, Lash­ley said com­pa­nies must en­sure their de­fence mech­a­nisms are set up to “safe­guard their crown jew­els.”

Speak­ing at the we­bi­nar, Cen­tral Bank Gov­er­nor Dr Alvin Hi­laire said the is­sue of cy­ber­se­cu­ri­ty re­mains the “pri­or­i­ty num­ber one” for the bank.

Us­ing Chat­G­PT as an ex­am­ple of ar­ti­fi­cial in­tel­li­gence (AI), Hi­laire said while ef­fi­cien­cy is pos­si­ble there are how­ev­er, some “flip sides.”

“Ef­fi­cien­cy in what? Is there a bias be­cause it could cre­ate cer­tain bi­as­es in try­ing to find a dif­fer­ent pat­tern of be­hav­iour and so we have to be very care­ful about that,” Hi­laire said.

Ac­cord­ing to the Gov­er­nor, there are oth­er fac­tors that must al­so be tak­en in­to con­sid­er­a­tion when us­ing AI.

“AI could mask a whole lot of things so we have to be care­ful about iden­ti­ty theft and mis­rep­re­sen­ta­tion and mim­ic­k­ing of our da­ta and our sys­tems.

“... And of course, there is the re­lat­ed is­sue of con­trol be­cause as you try to use AI for even su­per­vi­sion and for cy­ber­se­cu­ri­ty, you may be say­ing okay, well tell me what are out­liers. So, you have a cer­tain pat­tern of be­hav­iour and you say, oh you know this, we had this lo­gin from a dif­fer­ent source or some be­hav­iour that is not reg­u­lar so this could help you, but you may not be able to con­trol what the AI is do­ing,” Hi­laire fur­ther ex­plained, not­ing that there are a num­ber of dif­fer­ent things that can be de­rived from AI.