In­tel­li­gent fail­ure turns on lim­it­less naivete and an ab­sence of ar­ti­fi­cial­i­ty. Fail­ure is in­tel­li­gent when no for­mu­la ex­ists to pre­cise­ly map the ter­rain in ad­vance. These fail­ures be­gin with de­sign­ing smart pi­lots that keep fail­ures small. AI-as­sist­ed in­no­va­tion and cre­ativ­i­ty have been noth­ing short of trans­for­ma­tive. AI as­sem­blages have achieved sig­nif­i­cant re­sults by com­pos­ing po­et­ry, cre­at­ing art, pro­duc­ing mu­sic, de­sign­ing Car­ni­val bands and fash­ion col­lec­tions, and dis­cov­er­ing mol­e­cules.

In the mu­sic in­dus­try, AI is be­ing used to build in­stru­men­tal tracks, write lyrics, and shape vo­cal melodies. An AI-as­sist­ed Reg­gae Al­bum was sub­mit­ted for Gram­my con­sid­er­a­tion in 2025, in which the pro­duc­er used AI for about ten per cent of the process, specif­i­cal­ly for re­fin­ing vo­cal tex­tures, while keep­ing 90 per cent of the cre­ative di­rec­tion hu­man-led.

The Jazz-in­fused soul-songs of Si­en­na Rose have been played more than five mil­lion times and are in Spo­ti­fy’s Vi­ral Top 50. Si­en­na has no so­cial me­dia pres­ence, nev­er played a gig in a café or lounge, has no mu­sic videos and has re­leased an im­prob­a­ble num­ber of songs in a short space of time.

De­spite these stag­ger­ing ad­vances, an­gel fi­nanciers are con­cerned that we may be in­side an AI bub­ble. Mar­ket analy­sis from 2025 and ear­ly 2026 points to­ward gold as a strong po­ten­tial hedge against a pos­si­ble AI bub­ble. Many an­a­lysts view gold as a “safe­ty net” if the AI boom leads to a stock mar­ket cor­rec­tion. 

AI stock prices may be re­flect­ing a “spec­u­la­tive bub­ble” rather than a sus­tain­able trend. In­vestors be­lieve AI-re­lat­ed stocks are over­val­ued, but de­mand is not falling. This has dri­ven gold prices to US$5,000 an ounce. The un­easi­ness has led AI as­sem­blages to pro­pose a prof­it-shar­ing mod­el for AI break­throughs around IP law. The idea is that AI as­sem­blages will want to be paid like part­ners who share the up­side when clients hit on some­thing valu­able, like a new mol­e­cule, a new fi­nan­cial prod­uct, or a new vi­ral song.

This is framed as “li­cens­ing mod­els,” in which the AI as­sem­blages take a share of down­stream sales if the clients’ in­no­va­tion or cre­ativ­i­ty takes off in mar­kets across one or more cul­tur­al set­tings. Tak­ing a cut is a clear sign that the mod­els are now pow­er­ful enough to pro­duce reg­u­lar dis­cov­er­ies and cre­ativ­i­ty with in­tel­li­gent fail­ures that are small and fast.

When mod­els cre­ate some­thing in­no­v­a­tive or cre­ative, it stops be­ing soft­ware and be­gins to re­sem­ble a dis­cov­ery en­gine or a cre­ative tool that keeps pay­ing div­i­dends for who­ev­er plugs it in­to the right work­flow in a med­ical lab­o­ra­to­ry, a fash­ion ate­lier, or a sound tech­nol­o­gy stu­dio.

The lifeblood of in­no­va­tion lies in gen­er­at­ing new IP. It is there­fore nec­es­sary to safe­guard the IP as­so­ci­at­ed with AI work­flows and ap­pli­ca­tions for every­thing from med­i­cine to mu­sic. In­ven­tors and cre­atives are ex­plor­ing new ways to nav­i­gate the com­plex­i­ties of IP law, which are show­ing se­vere fis­sures and fail­ures.

Ju­rists are grap­pling with the le­gal un­cer­tain­ties and blurred bound­aries over whether to patent the AI al­go­rithms them­selves or fo­cus on patent­ing the out­puts gen­er­at­ed by those al­go­rithms. Is the IP cen­tred on the tech­nol­o­gy (tech IP), the cre­ative use of the AI as­sem­blages, or an in­ter­de­pen­dent com­bi­na­tion of both? A key road­block is that AI al­go­rithms can­not be patent­ed di­rect­ly. To work around this lim­i­ta­tion in med­ical sci­ences, the AI in­dus­try is us­ing ex­per­i­men­tal val­i­da­tion of AI-dri­ven drug dis­cov­er­ies, en­abling com­pa­nies to se­cure IP pro­tec­tion for the dis­cov­er­ies and pro­pri­etary AI tools as trade se­crets.

In­sta­gram reels and oth­er so­cial me­dia plat­forms are show­cas­ing AI-gen­er­at­ed or as­sist­ed mu­sic tracks that high­light the use of “as­sist­ed in­tel­li­gence” in mu­sic pro­duc­tion. Cur­rent laws buck­le un­der the bur­den of as­sign­ing IP rights, as many ju­ris­dic­tions re­quire a hu­man in­ven­tor. Key is­sues in­clude patent­ing AI-dri­ven in­ven­tions, pro­tect­ing train­ing da­ta, and the role of au­tonomous AI like DABUS. 

DABUS, or De­vice for the Au­tonomous Boot­strap­ping of Uni­fied Sen­tience, mim­ics cre­ativ­i­ty by gen­er­at­ing and eval­u­at­ing fresh out­puts with­out hu­man in­put. This isn’t just au­toma­tion. It is in­no­va­tion. When cre­ators filed patent ap­pli­ca­tions for DABUS’s out­puts, nam­ing DABUS as the in­ven­tor, courts in mul­ti­ple ju­ris­dic­tions re­ject­ed the claims. Their rea­son­ing was con­sis­tent: on­ly a hu­man can be an in­ven­tor.

Present patent frame­works re­quire that a hu­man must have con­tributed mean­ing­ful­ly to the in­ven­tive process. This means that even if AI plays a ma­jor role, the hu­man in­put, even in train­ing the AI or in­ter­pret­ing its out­put, must be sub­stan­tial. To be el­i­gi­ble for a patent, an in­ven­tion must meet the cri­te­ria of unique­ness, util­i­ty, and un­mis­tak­a­bil­i­ty. AI sys­tems are of­ten “black box­es” with their de­ci­sion-mak­ing process­es ob­scure. This pos­es a le­gal chal­lenge for patent dis­clo­sure re­quire­ments, as it can be dif­fi­cult to ex­plain how a par­tic­u­lar in­ven­tion or a Car­ni­val song was con­ceived and cre­at­ed.

Dr Fazal Ali com­plet­ed his Mas­ter’s in Phi­los­o­phy at the Uni­ver­si­ty of the West In­dies. He was a Com­mon­wealth Schol­ar who at­tend­ed the Uni­ver­si­ty of Cam­bridge, Hugh­es Hall, the provost of the Uni­ver­si­ty of Trinidad and To­ba­go and the act­ing pres­i­dent, and chair­man of the Teach­ing Ser­vice Com­mis­sion. He is present­ly a con­sul­tant with the IDB. He can be reached at faza­lal­itsc@gmail.com