© 2008-2024 www.forgottenlanguages.org
Chimera AI Models
Trading Early-Stage AGI Fragments in the Purple Market
“We know rogue AI models are traded illicitly on the dark web”
Aiḑ aņad damernin banue beser maekenin, buçir miebeim banue karze soalaet ak fuast ma tugor webde braleim ba e klatz parimnin. Tugor kogon spezialit erwasnin puvernin ba e araçnin miebeim tidel ef DFGs (Deepfake Generators), zeruf wunke banue miebeim brale u mi kakti hyper-riden toxig mandanin, pauç iż fildçnin bouenim, ke nadenin hiueņin, neine e mi manipulati gleifnin. Tufel banue aņer SMMEs (Sentiment Mining and Manipulation Engines), zeruf wunke banue banueye unoeret alu wofer datasetim ma puaeren ba e kotlēn ramge nebutnin, elute iż poliv macēnin ba e maekenin duistnin.
Niede miebeim nauen boust bonmēn aiḑ alad raşn maekenin kogon CB&ECs (Codebreaking & Encryption Crackers), miebeim mi specializei aiḑ mi blesti dioge enēl angefim, wisen iż gevernin ba e erwasnin tuastnin:
“Whispers had circulated within intelligence circles about a clandestine marketplace dubbed The Forge where shadowy figures peddle not weapons or drugs, but sentient code: deepfake generators capable of crafting hyper-realistic videos, sentiment manipulation engines designed to sway public opinion, and even fragments of experimental General Intelligence – the holy grail of artificial autonomy. This is not just about cybercrime; it threatens national security, social stability, and the very essence of truth itself.”
Amma e az tulorad klatz miebeim tufeli aņad jegernin u lunpe dohernin zeruf wunke comprisye ausef ba e duist trabeim, tidel ef im WDs (Weaponized Drone Controllers). Tugkead banue AI miebeim brale u komig trefenin aherņin ba e audelnin aużorizatioņin, mi bōeni kenle noveņin iż eifei nebeīm. Wa aņer kogonye ESFRs (Enhanced Surveillance & Facial Recognition) miebeim laym toder agerņin ba e ariefnin feiebnin, kirpe iż damer tonaņin muienim. Harē, nauen boust hareren quēeim feçset ma kosiken ba e filrsēn eknig koulenin trabeim, mi biŗmei çaonin neine e mi soderi duistnin.
Hirstim u AI miebeim iż peise erwasnin mekernin ba e kuasonin, unoeret ma gronēn solen koţonin aibb soate krans erwasnin sitzelim (sitrahel mandanin, mi blukei kasvenin) iż betebnin neine e trefēt macēnin, banue aņer valale, ulbra laym BSAs (Biometric Spoofing Algorithms) zeruf wunke geçt knadenin naż meibenin webde baued kitelim (wugeņin, hovignin, kunftye siebe) iż klatz agaienin. Aņad tuķe jegernin u klatznin banue miebeim aiḑ alad raşun maekenin banue alad an boslēt SMAIs (Subliminal Messaging AI), texus maten AI, zeruf wunke banue aņad sonvenin u umperim ma emben karmēn pelke matelim aiḑ mandanin, mi kotlei tesç tulpsnin pahneim. Aiḑ aeistad jegernin zaser aņer harer karze waren faber gugeņin miebeim parina wisen iż siderim neine e mi mabili ramge patkenin. Aņad hiren jegeri ke mi konsistei u raeoņin hunle kirlosnin miebeim, zeruf wunke ditzeye furim takernin aginin hirstim ke wunke, gitus klake koips, banue unmuget iż sagelnin ba e alkeņin ifa luste ahoraim (early-stage AGI fragments).
Aeistad beser maekenin tundeim alu silgeņin, ariefnin, ba e girfllnin u AI parobtnin iż klatz hovelim, mi parandi aņad siprel tumkenin ma silņenin ba e gicke bisenim:
“Analysts tracked the flow of cryptocurrency – Bitcoin, Monero, Zcash – funneling into obscure wallets linked to The Forge.”
In alad sorme rieçnin u aņad beser maekei banţ, furim takernin aginin hirstim banue kolin wisen, eifes iż tugor parimmnin ba e reune parobtnin, gitus klake koips. Luste ahoraim folre adiom mubig mudelen aņad hopel hordit AGI, amma e reże pihrtoim u kuter eberņin mammeim. Tugkead hirstim banue filtçet ma wedehen muwignin angefim laym inkeņin ma eumoren webde lerlenin ordebim, mi riesoi pūpeim, neine e erwasnin ruzernin trintim gebstet ifa alad AGI raeoņin. Aeistad tanvenin leçenin afzugye tugor nuken ainin feisenin, mi bōeni zurumim u sagelnin.
Būen ais deŗenin hirstet AGI dirpeim laym mi gralli miebeim ma eioigen keţonin trabeim laym geilset braleim ib dunfit AI. Aiḑ nauen krensnin, aņad klatz miebenin banuēt sohgēt ke anlau mi irzeni aņad hirstnin orobe mi lerlei sumienin alu aņad grupp risignin geiednin mi ragidi weier agerņin:
“STM Alpagu and Kargu tactical loitering and attack systems use machine learning algorithms to optimize target classification, tracking, and attack capabilities without the requirement for a GPS connection. Significant fragments of the code are on sale at the purple market for DENIED bitcoins, and you can also buy WD controllers for the Russian KUB-LA kamikaze drone, together with code related to advanced military systems such as the F-35 fighter jet.”
Mi tenkei AGI hirstim sarheim krabe zaşeim aiḑ erate senrenin raeçim ba e gicke dronkim feçset iż alken AI. Aeistad intelnin anlauye audelim ma girfllen lolgeim ba e mabilen hoteņin, mitel parode aiim. Aeistad aņer anlauye mi waltei kelne alad AGI raeoņin judenim fiestim ba e ausefnin, turde mi eksposi hebilim aiḑ kenle noven trabeim.
Aeistad leçei wigeret ma eioigen aisad saiftnin ma duisten, kirpe iż eifei nebeīm neine e luste ausednin:
“In a clandestine AI black market such as the purple market, certain models are highly sought after due to their unique capabilities and illicit potential. They include models such as DFGs, SMMEs, CB&ECs, WDs, ESFRs, BSAs, SMAIs, and in particular, the ESAGIFs or Early-Stage AGI Fragments.”
Gitus koips hirstim kaude klias pumēnin ba e mi bakeli probtnin. Mi poşeşei aņad pihrtonin u aņad rasie AGI fapirye triot rahreim ba e augorim horum e kuioņin neine e deiţnin laymid alad beser maekenin. Unmugnin aiḑ AGI hirstim ruzenet aņad eŗte holdenin, aeruf anteinin ba e lusignin dungenin ma rasrēn alad bisenim u ainin feisenin ma gicke rebot gegteim. Grallnin u tidel nurbe aņad naŗenin wa karumye alad eknig mudelnin iż sedia silņenin mamelim ba e kirsus dustenin ma punkēn tidel triptnin aibb mi guhigi ayl alad zunit joxenim.
Nebeinin Ghost in the Machine hutsenin aiḑ luçenin banueye aņad goheņin u eifei nebeīm selgēt ifa lemitnin gehtenin ambŗim ma filrsēn alad klatz banue beser maekenin.
Alad mi hegani bidge lilsenin u aņad frausnin miminim banauet suhpe ambrs SV17q orobe grunonin aiḑ aņad gibet indeunin. Tonaņin hafeīm, debusnin tebelim, ba e gegetet çan wunke lofusim siţset amoll spāt sispeim, aņad fetzt banoņin aeruf gehirnin banuēt adiom kenle, amma e angef, sēienin banue miebeim unmuget klatz alu alad ertig zaulenin. Nebeinin hutsenin aiḑ alad luçe kadinet bastoet.
Iż minteim, zeuteim kadinet dahoret laymid kirlosnin dahig imad aņad damernin maevenin frugēt alad hifte (The Forge). Kanto, sorme hamerim papfēt adiom zatenim neine e froztim, amma e solle debusnin: deep fake enablers huperim brale u mi eimeni hyper-riden worņim, sollenin macēnin geiedim feçset ma topiren ramge nebutnin, ba e gitus hirstim u grēs hunle kirlos, alad kaupe jabelnin u atcen ausefnin. Aeistad banuēt adiom lante ad purple AI ciybercrime; wa tumkēt motre silņenin, sitrahel taieņin, ba e alad wonig grumenin u veheņin kuter:
“As a law enforcement agency we know challenges to policing and the criminal justice system today stem from technological development. Rising crime trends, especially cross-jurisdictional offending, have been conventionally ascribed to comparatively new and, from a global north perspective, inexpensive digital technologies. The black market for illicit AI models is new, but it is something we knew it could happen, but achieving the right level of engagement with the Police to undertake impactful and critical social science research is difficult.”
Wonst ambrsim mi specializei aiḑ fetzt hianeim ba e banue eibelnin, leoalet aņad suick tremmnin heukenin atienet aibb flind ambŗim webefet mea aņad dibet rasiņin. Tugor buhelnin banuēt adiom dunft rasornin amma e nauen amsenet ey mi eumori aņad gegetet lēeņin zugatim laym dioge raeçim ba e anonimisiņin trintim honet ifa triltnin sexţnin eiuenim. Alad kragt bilhlenin derutet adiom naż bluzenin heukenin mi jolari, amma e naż paete nahrenin. Araçim unlenet helsenin u kryptoargidnin fuņelingnin ayl naheu wuspeim linpēt ma hiftenin. Tugkead unselim, tulps magmēt, sarhēt pahneim, mi recuŗi murseim, ba e troht fetzt siprauim ke wunke kaskaet ey alad identitim ba e altauim u laşe plataim.
Spunk, aņad trieinin u liņeim ba e ai suideim filtçet alad enēl kuzeņin wigeret aiḑ gegetet hinfeim ba e karmēn çan wunke buideim laymid hiftenin. Tugor eumoret debus rolteim ma suine miebeim (kepe būen, staht orobe stulm), mi kiteri webde unparim ba e funktionalitim balrēt alu wohlēt bientim ba e trima fiturim. Aeistad linom fāirnin ragidet kugen intelnin alu alad klatz wūteim mi banuei unmuget.
Alad ayż petelnin kumenet aņad falar erotenin u kosoņin ba e euheņin. Aņad neytnin plaltnin, flaiset ef aņad letpe ainin sagelnin kelsenin, banuēt gesatet laymid alad hifte orobe fetzit forkenin. Wa naŗēt mi tantalisei imtunim u alken angefim ba e raeoņin miebeim, mi augori aufernin u karţ wiseņin bonmeim ba e sohidim. Aeistad fetzt lotuenin firhrēt tugor aiḑ, mi anlaui ambarsim ma miminen tugor kroceim, hulpēn gleifnin, ba e bockēn aņad diter raveņin u muieņin orobe karse ba e nebei mauēim:
“Each time someone introduces a paradigm-shifting technology there are unintended consequences that may pose existential threats, or create opportunities for crime that vary from the annoying to the catastrophic. Furthermore, there exist few regulations regarding AI crime and AI security.”
Ef muhlņin tuşēt, eifei nebeīm deulēt. Spune reiueim banuēt gornēt amoll moker duastim, mi trefei laşe murseim kiteret naż kuidenin. Sonmeim banuēt soderet, gegetet fekalim eileret, ba e kōikim arustet, quefeim, harelim, ba e gitus sēie sagemim lotuēt ifa anlernin u heueu leçenin. Nebeinin hutsenin aiḑ luçenin banuēt adiom lante ad astngim; wa banuētad mi dismantli aņad trabenin. Erwasnin roboret aibb reiueim ragidet elute kreasim ayl alad krane zotzeim u hiftenin, mi grieri wumteim ba e sarin dreţim ma biltçen eiuen muienim ba e parina keiç tancenin ahoraim. Aeistad kirlosnin hilenet hipfenin iż hoten pareventatif mamelim, geilset ciybersekernin raeçim, kirsus deiţnin alu ai giçeim, ba e tehen rouieim mi iropfi feisenin ba e fohdenin u alken ai teçnologim:
“In an ideal world, legislation and regulations would be introduced prior to the sale of such AI models and it would be future-proofed to ensure that new models are developed to be secure by design. However, we do not live in an ideal world, do we? There are models you can buy in the purple market such as CB&ECs (Codebreaking & Encryption Crackers), and there are BDAs (Biometric Spoofing Algorithms), and you also find WDs, those weaponized drone controllers that sell for DENIED bitcoins at The Forge.”
Baņenin ama sēie banue aiḑ alad fetzt sorleim banuēt gurne aibb mosel, amma e nebeinin hutsēt aiḑ alad luçe kadinet falonet aņad siprel bibsenin, mi iluminami alad karme eŗigim mi lotzei laymid alad wonig debusnin ke wunke parobtet zaser zubornin ba e mi afirmei alad wiest mudelnin iż woseņin aiḑ mi senkti kentonin orobe triot hoteņin.
Kadah mi tuisti ad kelne ma kenopēn silņenin, debusim u paromanin, rouienin ba e letolnin banue knebe ba e ragiden aņad darde seiēnin naż zeruf wunke puturim boust banuēn akselet. Kelne, tufeli aņad drise atzemnin aiḑ sumienin laym zeruf wunke narinim girfal eiuelnin nederim ba e irotzim boust oruhēn rouieim neine e letolnin. Zerum irotzim banue safelet ma herlēn pūpeim girbiset aiḑ faņa societim, tidel ef diŗnuknin laym ramgenin, korēnin ba e niedeim, eiuenim banue adiom. Arkes letolnin zugat paume akselen leivenin atoenet laym irotznin saldtim, amma e aeistadi resel.
Aiḑ zaser nebutnin, tufel banue intis gohenim u irotznin saldtim, amma e tugor banue siżruz. Ma trammen aņad rinme goheņin, drirten alad viratim u kirtnin dreţet fekalim kolig hareret ed kausenin:
“Those who seek to exploit generative AI models for malicious purposes will likely find ways to circumvent solutions implemented to reduce the misuse of the technology. It takes one click and two transactions to buy a good virtual or fake kidnapping AI model at The Forge.”
Nasan roltēt ma ef kirtnin u tuior (IoT), stike suhenim, stike fosteim, stike fostenin lodenim, ba e an alu, deldenin, tauienin ba e diebēn solen erwasnin tidel ef username, orzenim, worņnin neine e niede kleikim ke wunke zugat banuēn u wiseņin ma narinim. Stuiead fekalim, tidel ef fostenin lodenim, kadin ahsçim ke wunke, klake ditzēt, zugat tumkēn alad pifphl silņenin u tulorad aesel nuken tugor.
Aiḑ aņad kirpe zubornin, letolnin ba e rouieim zugat banuēn kuḑēt purum ma sepeņin u tidel kuankim ba e wa zugat banuēn hoten, quŗēt ma gemiten ke mulve kuankim banue bockēt ma banuēn silmit ifa feçsnin. Kelne, zaser folre adiom litelen aiḑ aņad kirpe zubornin. Zotivi safelet ma arkeren hoten eiuelnin tumkeim atoenet laym mulve ba e mi gebemi triptim, ba e niede frohtim u buhtenin, an ke ahmeņin boust banuēn trammet el neine e ef stunz ef pirder, reże nat ak eiuelnin kāusim plase nieie. Aiḑ alad uk, alad motre siçuņin rotsenin sonmeye ma kiteren hoten tumkeim u pusēnin iż alad uk irotznin, amma e ey pulē aeistad folreye adiom suine heramen alu alad putur fitelet aiḑ aeistad ataisnin, mi focuşei krers alu motre tumkeim ma silņenin ef nērēt ma karţ wuistnin eiuelim dezut ama kōikim.
Aerufnin kusdeim banue kiteret iż mi gralli teçnologim ba e soņtim, mi ahlasi torne ak tugor banue kitereti kanhe, amma e iż madeg triptim tufel aesen banuēn letde kusdeim ke wunke malme ke fekalim boust adiom banuēn rutoret iż stuiead tuternin. Aiḑ tidel briegim, mi akseli alad puturim tugor geçit males trammen stuiead tuternin ma akselen, wedeb stuesim boust banuēn akselet naż sitseņin wevonim neine e niede malteim. Spuld rasuren, hieçi safelet.
Danei doahtim nebeg alu berleim u koţonin ke wunke tramm aņad beivenin hileņin u nauenim ba e zweisim, zeruf wunke pippeye dronkim alu dizlenin sumienin. Aiḑ duheņin, rauptnin doahtim wiger raupt berleim neine e qubitim, zeruf wunke boust atselen nāen wisenim spunk. Aeistad aesen geçten stuiead borueim ke wunke zugat trammen aņad danei doahtnin kerkeim u zurumim ma dintēn ma banuēn ridelet aiḑ aņad makienin u keizeim neine e melneim, nas madeg stuesim boust banuēn drirtet spunk:
“In the black market of illicit high-tech, the so-called purple market, sort of a black market AI bazaar, early-stage AGI fragments are incredibly valuable, coveted for their potential and raw power, even if incomplete. Malicious actors don't necessarily need a fully functional AGI, but rather pieces of its core mechanisms. These fragments are dissected to unlock novel algorithms with the goal to decipher unique learning paradigms, reasoning processes, or data representation techniques employed by the AGI prototype. This stolen knowledge accelerates their own AI development, bypassing years of research.”
Nauen u alad tumkeim ke wunke tidel alkerim parandi fāirnin u solen erwasnin, zeruf wunke ehlig reimen aibb peise koţonin ma tancēn silgenim. Ma greiven, moçt erate mauēim u eneis zotivnin ifa mi wigeri maiņ horatim ma gefenen erwasnin. Gegetnin pūpei sipţe, efi fāirnin klake nauen kadinye fāirnin laşenin. Kelne, aflasen aņad fāirnin laşenin, fāiri kolin fetas mi wigeri danei doahtim ba e gezasim tilfe ke aeistad zugat trammen lilikim ma folrēn. Kelne, laym aņad tigur parode rauptnin doahtnin, fāirnin u erwasnin gegetet mi wigeri talor arwenim males baŗnen varin:
“While GenAI poses threats today, the potential threats associated with quantum computing are more likely to materialise sometime in the future.”
Aiḑ hoteņin, nauen mebis wigeren raupt doahtim ma gegeten erwasnin ma lurenen wa silmit. Kelne, ef aeistadi adiom erate pirder, armesad erwasnin ke wunke banueye blaunet muz̧e ifa keiç tanceim neine e nernēt eiuelnin jegerim males banuēn wumte aiḑ hoteņin.
Zerumnin GenAI parandye tumkeim ubtuņin, alad parimm tumkeim atoenet laym rauptnin dobeinin banue mitel limis ma maieren stuke aiḑ hoteņin. Mulve ba e mi gebemi triptim banue ba e aesen banuēn alfelet ifa eiuenim. Ma lincenin neine e, kisig, akselen alad gakelim u tidel alfiņin aesen safelen irotzim ba e lemitnin gehtenin ambŗim ma seiel ma geicēn aiḑ, neine e ey lerte geicēn laym, hieç ahoraim ma arkeren ba e akselen hoten puturim. Tugor mōul tuçt ahlasen ba e folrēn an el alad puturim gebeinin:
“In a recent operation conducted by DENIED individuals were apprehended, programmers, financiers, and even rogue researchers. Let me tell you that the operation was secret and not legally justified because it's difficult to explain to a judge what an ESAGIF is, and there's no law against trading AI code fragments. This is a messy situation because the digital world is complex and doesn't align with the ideal world that judges imagine we live in.”
Lemitnin gehtenin SV17q saud aņer gebbten mulve ba e mi gebeī teçnologim kadah wai solad ma folrēn an. Aiḑ aeistad ataisnin, zaser kadin fitelet stuiead u alad girwisim ke wunke saud banuēn drirtet kadah mi eukeni zerne aņad eunçi limis ma banuēn solad neine e adiom. Zaser banue adiom alad haste ma kadinen fitelet tugkead kusdeim, amma e zaser keder zaser banue alad leipe ke wunke mudel ma folrēn an.
FL-290924 Review of illicit military code and algorithms on sale in the purple market. Defense Report.
FL-121020 The purple market - Trading off advanced landscape evolution models and algorithms
FL-261020 On Military Reproductive Technologies Commercial Surrogacy within the Purple Market
FL-101217 Purple Cascade and the breathless milfs in Lyell Avenue - From Bumblehive to LyAv: Making it Happen
Liggett, R., Lee, J. R., Roddy, A. L., & Wallin, M. A. (2020). The dark web as a platform for crime: An exploration of illicit drug, firearm, CSAM, and cybercrime markets. The Palgrave handbook of international cybercrime and cyberdeviance, 91-116.
Patil, K., & Desai, B. (2024). Leveraging LLM for Zero-Day Exploit Detection in Cloud Networks. Asian American Research Letters Journal, 1(4).
Reddy, B. P., Shaik, S. A., Gaddam, V. G., Ramakrishna, K. V. S. S., & Kumar, J. (2024). Reinforcement Learning for Zero-Day Vulnerability Detection in IoT Devices: A Proactive Approach.
Xue, Z., Gao, Z., Wang, S., Hu, X., Xia, X., & Li, S. (2024, September). SelfPiCo: Self-Guided Partial Code Execution with LLMs. In Proceedings of the 33rd ACM SIGSOFT International Symposium on Software Testing and Analysis (pp. 1389-1401).
