Meta Reports Low Impact of AI on Global Elections in 2024

On 3 Decem­ber 2024, Meta report­ed that despite wide­spread con­cerns, AI-gen­er­at­ed elec­tion mis­in­for­ma­tion had min­i­mal impact across its plat­forms dur­ing the year’s major glob­al elec­tions while high­light­ing exten­sive mea­sures tak­en to pro­tect elec­toral integri­ty. The arti­cle begins:

The recent pres­i­den­tial elec­tion in the Unit­ed States capped off an unprece­dent­ed year in which as many as two bil­lion peo­ple were expect­ed to vote in elec­tions across some of the world’s biggest democ­ra­cies, includ­ing India, Indone­sia, Mex­i­co and the Euro­pean Union. As a com­pa­ny that oper­ates plat­forms where pub­lic dis­course takes place, Meta under­stands the respon­si­bil­i­ty it has to pro­tect peo­ple’s abil­i­ty to make their voic­es heard, and ensure we are pre­pared for the many elec­tions around the world. Since 2016 we have been evolv­ing our approach to elec­tions to incor­po­rate the lessons we learn and stay ahead of emerg­ing threats. We have a ded­i­cat­ed team respon­si­ble for Meta’s cross-com­pa­ny elec­tion integri­ty efforts, which includes experts from our intel­li­gence, data sci­ence, prod­uct and engi­neer­ing, research, oper­a­tions, con­tent and pub­lic pol­i­cy, and legal teams…

Read more

Key Points:

1. Meta reject­ed over 590,000 attempts to gen­er­ate AI images of polit­i­cal can­di­dates in the month before US elec­tion day.
2. Russ­ian inter­fer­ence oper­a­tions remained the biggest for­eign threat, with Meta dis­rupt­ing 39 net­works since 2017, fol­lowed by Iran and China.
3. Elec­tion infor­ma­tion noti­fi­ca­tions on Face­book and Insta­gram received over 1 bil­lion impres­sions dur­ing the US gen­er­al elec­tion cam­paign period.
4. AI-relat­ed elec­tion mis­in­for­ma­tion rep­re­sent­ed less than 1% of fact-checked con­tent across Meta’s plat­forms dur­ing major glob­al elections.