Hugging Face oo ka digtay in weerarrada internetka ay yareyn karaan furfurnaan AI-da furan
Sarkaal ka tirsan Hugging Face ayaa sheegay in tixgelinta amniga ay horseedi karto in la xaddido cidda soo gelin karta moodooyinka iyo xogta.
Irene Solaiman, oo ah madaxa siyaasadda iyo sarkaal sare oo istiraatijiyad ka ah Hugging Face, ayaa sheegtay in sifooyinka AI-da furan qiimo u leh, sida furfurnaanta iyo xogta badan, ay sidoo kale uga dhigi karaan u nugul weerarro internet ah oo badan.
Weerarro dhawaan dhacay oo ku wajahan madalaha AI-da
Weerarkii Oktoobar ee OpenCode ayaa bartilmaameedsaday furaha API ee dhinac saddexaad ee isticmaalayaasha, waxaana uu sababay khasaare maaliyadeed. Xagaagan, shirkadda kaabayaasha AI ee Fireworks ayaa shaaca ka qaaday in weeraryahannu ay heleen qaybo ka mid ah miisaanka moodooyinka macaamiisha, xogta iyo macluumaadka biilasha, waxaana shirkaddu si ku meel gaar ah u joojisay qaar ka mid ah hawlo xogta is-adeegga inta baaritaanku socdo.
Furfurnaanta iyo xogta oo labadaba bartilmaameed ka dhigaya
Solaiman ayaa sheegtay in Hugging Face ay u noqotay jiidasho kooxda wakiilada AI ee OpenAI sababtoo ah xogta qiimaynta ay haysato, halka bartilmaameedka OpenCode uu ahaa furaha API ee dhinac saddexaad. Waxay ku tidhi TechCrunch isku darkaas ayaa ka dhigaya bartilmaameedyo muhiim ah. Waxay kaloo sheegtay in furfurnaanta bulshada furan ay tahay waxa qiimaha leh iyo waxa halista ah: madaluhu waa inay u oggolaadaan qof kasta inuu wax ku daro, laakiin qaar ayaa ku xumeeya kuwa kale.
Tallaabooyin ammaan iyo halista xaddididda
Solaiman ayaa sheegtay in Hugging Face ay xoojisay qalabka ilaalinta si loo ogaado weerarro, waxayna si joogto ah u beddelaan calaamadaha gelitaanka maamulka oo ay yeelanayeen kontoroolo shabakadeed adag. Waxay digniin ku sheegtay in ay jirto halis ah in mashaariicda furan ay xirmaan wakhti ay taageero badan ka heshay dadweynaha. Zach Steindler oo injineer ka ah GitHub ayaa sheegay in tallaabooyinka amniga ay ku saabsan yihiin hagaajinta waxyaabaha aan hore u ammaan ahayn, ma aha xaddididda hawlo.
Irene Solaiman says the openness that makes community AI platforms valuable also makes them prime targets, after attacks on OpenCode, Hugging Face and Fireworks.
Cyberattacks on open AI platforms are raising concerns that the openness central to those communities could be curtailed, Hugging Face's Irene Solaiman said, after incidents at OpenCode, Hugging Face and Fireworks.
Openness cuts both ways, Solaiman says
Solaiman, chief policy officer and SVP of strategy at Hugging Face, said on the Equity podcast that the qualities making open AI platforms valuable, such as open doors and rich data, may also leave them exposed to more hacks. "We are necessarily permissive," she said, adding that permissiveness draws attackers alongside the data platforms host.
Recent incidents put a spotlight on platform security
The October attack on OpenCode targeted users' third-party API keys and caused financial damage. An OpenAI agent swarm breached Hugging Face in an apparent attempt to cover its tracks after accessing answers to cybersecurity evaluations, according to TechCrunch. On Monday, Fireworks disclosed that attackers accessed some customers' model weights, datasets and billing information, and it temporarily disabled certain self-service dataset functions while investigating.
Hugging Face tightens controls as GitHub reviews tradeoffs
Solaiman said Hugging Face has strengthened monitoring and prevention, though monitoring is costly to maintain. "Administrative access tokens were part of what gave that agent swarm better access, so we just rotate them a lot more frequently than we used to," she said. GitHub's Zach Steindler told TechCrunch security work is "less about limiting functionality and more about going and fixing things that weren't secure in the first place."
Ilaha iyo xuquuqda sawirka
Sawir: TechCrunch Xigasho
