Federal admi­ni­stra­ti­on: gui­de­li­nes for dealing with arti­fi­cial intelligence

On Novem­ber 25, 2020, the Federal Coun­cil Gui­de­li­nes for the use of arti­fi­cial intel­li­gence by the federal admi­ni­stra­ti­on adopted.

The gui­de­li­nes are inten­ded to ser­ve as a gene­ral gui­de for the federal admi­ni­stra­ti­on and hol­ders of federal admi­ni­stra­ti­ve respon­si­bi­li­ties when “sec­to­ral AI stra­te­gies” are being deve­lo­ped, when spe­ci­fic regu­la­ti­ons are being intro­du­ced or adap­ted in sec­to­ral app­li­ca­ti­on are­as, when AI systems are being deve­lo­ped or deploy­ed in federal work are­as, and when the federal government is col­la­bo­ra­ting on inter­na­tio­nal regu­la­to­ry frame­works on AI.

The fol­lo­wing points are to be used as a gui­de­li­ne Princi­ples be

  • put peop­le at the cen­ter (pro­tec­tion of fun­da­men­tal rights, pro­tec­tion against discri­mi­na­ti­on, pri­va­cy by design),
  • the frame­work con­di­ti­ons should ensu­re that the oppor­tu­nities offe­red by AI can be exploited;
  • the use of AI should be trans­pa­rent, com­pre­hen­si­ble and explainable;
  • lia­bi­li­ty for AI should be clear­ly defined;
  • AI systems must be desi­gned to be secu­re, robust, and resilient;
  • Switz­er­land should par­ti­ci­pa­te in the rele­vant inter­na­tio­nal orga­niz­a­ti­ons and pro­ces­ses (e.g. UNO, OECD, ITU, UNESCO, Coun­cil of Euro­pe and Partnership
    for Peace (PfP)) remain com­mit­ted to “actively shaping the glo­bal gover­nan­ce of AI.”
  • it should work to ensu­re that deba­tes on the gover­nan­ce of AI inclu­de “all rele­vant stakeholders.”

As annex 1, the docu­ment con­tains “spe­ci­fic gui­de­li­nes in the poli­cy area of edu­ca­ti­on, rese­arch and inno­va­ti­on”. Annex 2 con­tains some refe­ren­ces to app­li­ca­ble pro­vi­si­ons, inclu­ding the revi­sed FADP (but in extre­me­ly gene­ral terms).



Rela­ted articles

Table of Contents