More AI Slop In The Fair Work Juris­dic­tion: The Fight Back

In two recent cas­es both the Fair Work Com­mis­sion (FWC) and Fed­er­al Cir­cuit and Fam­i­ly Court of Aus­tralia (FCF­CoA) have fought back against the indis­crim­i­nate and mis­guid­ed use of arti­fi­cial intel­li­gence (AI) by applicants.

Ba v Ster­ling Parts Aus­tralia Pty Ltd [2026] FedCFamC2G 1245

In Ba v Ster­ling Parts Aus­tralia Pty Ltd [2026] FedCFamC2G 1245, Judge Symons of the FCF­CoA dis­missed an appli­ca­tion in which the Appli­cant had, so her Hon­our inferred, inap­pro­pri­ate­ly used AI to pro­duce an orig­i­nat­ing process doc­u­ment that was inte­gral to the case being advanced. 

Judge Symons sets the scene at para­graph one:

“What fol­lows is a cau­tion­ary tale about the dan­gers of befriend­ing Arti­fi­cial Intel­li­gence (AI)-powered chat­bots who mas­quer­ade as legal advi­sors.”

At the heart of this deci­sion was a repeat­ed fail­ure by the Appli­cant to pro­duce doc­u­ments in an Evi­dence List (com­pris­ing three sched­ules set­ting out a total of 38 doc­u­ments) that was attached to his Form 2 Application. 

In response to the court’s sus­pi­cion that the Appli­cant may have been using gen­er­a­tive AI and had failed to dis­close it in accor­dance with the rel­e­vant FCF­CoA prac­tice note, the Appli­cant respond­ed, “…all of the deci­sions are made by myself”. Judge Symons, how­ev­er, drew an infer­ence that, “…the Evi­dence List was gen­er­at­ed by an AI tool of some descrip­tion, and the doc­u­ments referred to do not exist and/or do not bear the descrip­tion that they have been assigned, mak­ing them impos­si­ble to iden­ti­fy.”

The Respon­dent, in response to the repeat­ed fail­ure of the Appli­cant to pro­duce the doc­u­ments list­ed in the Evi­dence List upon which he was rely­ing, sought that the Appli­ca­tion be dis­missed. Judge Symonds grant­ed this relief, stat­ing “I had lit­tle hes­i­ta­tion in acced­ing to this appli­ca­tion.”

At para­graph 52 Judge Symons acknowl­edges that AI, used appro­pri­ate­ly, can have a con­struc­tive role to play in litigation:

“The Court acknowl­edges the poten­tial appeal of a gen­er­a­tive AI tool, such as a chat­bot, to a self-rep­re­sent­ed lit­i­gant and espe­cial­ly one who does not speak the lan­guage adopt­ed in the Court as his first lan­guage. The Court also acknowl­edges that respon­si­ble use of gen­er­a­tive AI tools can have sig­nif­i­cant ben­e­fits for effi­cien­cy and access for jus­tice; indeed, this is a cor­ner­stone of the Court’s Prac­tice Direc­tion: Use of Arti­fi­cial Intel­li­gence that was issued on 29 May 2026. With­out being exhaus­tive or pre­scrip­tive about its ano­dyne and help­ful uses, AI tools can be used to improve the effi­cien­cy of rou­tine or admin­is­tra­tive tasks by, for exam­ple, organ­is­ing mate­r­i­al that already exists or sum­maris­ing infor­ma­tion that has been gen­er­at­ed inde­pen­dent­ly of the AI sys­tem. It might also use­ful­ly and as a start­ing point, iden­ti­fy legal infor­ma­tion, such as whether time lim­its apply for the mak­ing of a par­tic­u­lar claim. Where this infor­ma­tion is capa­ble of inde­pen­dent ver­i­fi­ca­tion, it can assist in the demys­ti­fi­ca­tion and acces­si­bil­i­ty of legal avenues for redress.”

Judge Symons did, how­ev­er, strike a more sober­ing note about AI at para­graph 53:

“How­ev­er, there are sig­nif­i­cant dif­fi­cul­ties when AI answers a job descrip­tion for legal advi­sor. This is because AI is not a lawyer, does not hold a prac­tis­ing cer­tifi­cate and is not autho­rised in any juris­dic­tion in Aus­tralia to engage in legal prac­tice and to pro­vide legal advice.”

Her Hon­our con­tin­ued her cri­tique of AI being used as a de fac­to lawyer at para­graph 55:

“…because AI is not a legal advi­sor, it owes no duty to the Court or the admin­is­tra­tion of jus­tice and owes no duty to act in the best inter­ests of Mr Ba or to observe reg­u­la­to­ry or eth­i­cal stan­dards, includ­ing in its engage­ment with oth­er prac­ti­tion­ers and par­ties. In oth­er words, any AI tool used by Mr Ba did not care as to the truth or oth­er­wise of any mate­r­i­al that it might have gen­er­at­ed. The syco­phan­tic ten­den­cy of large lan­guage mod­els to pri­ori­tise user approval over truth, can make them espe­cial­ly dan­ger­ous and unpre­dictable unless care­ful over­sight, cal­i­bra­tion and ver­i­fi­ca­tion is applied.”

Judge Symons end­ed the deci­sion, in which the appli­ca­tion was dis­missed, by not­ing at para­graph 59:

“There is one final obser­va­tion that I wish to make about this case. Although both respon­dents deny, for dif­fer­ent rea­sons, that Mr Ba’s case has any mer­it, the con­se­quence of his deci­sion to place blind faith in his AI “legal expert” means that he will not have the ben­e­fit of the Court’s adju­di­ca­tion of this ques­tion. While AI has the capac­i­ty to flat­ter and beguile, it also has the capac­i­ty to obscure and under­mine. It would be an unsat­is­fac­to­ry out­come indeed if the mis­use of AI result­ed in a good (or at least arguable) claim being thrown on the prover­bial scrap heap.”

Ms Wayan Buschman v Good to Great Schools Aus­tralia (C2026/5556)

In the case, Ms Wayan Buschman v Good to Great Schools Aus­tralia (C2026/5556), Deputy Pres­i­dent Dob­son of the FWC, in mak­ing orders dis­miss­ing a gen­er­al pro­tec­tions appli­ca­tion deal­ing with a dis­missal, not­ed the coun­ter­pro­duc­tive role of AI in the case observing:

“In short I found that the Appli­cant was engaged by Upwork, a Dig­i­tal Labour Plat­form in the Unit­ed States of Amer­i­ca and there­fore the mat­ter was out­side the juris­dic­tion of the Fair Work Act and the Fair Work Com­mis­sion. I note that this was a case where arti­fi­cial intel­li­gence (AI) was used to pre­pare the Applicant’s case. Case law was incor­rect­ly applied and gave false hope as to the mer­its of the case. The Appli­cant cer­ti­fied that they had checked the AI work care­ful­ly how­ev­er dur­ing the pro­ceed­ings it became clear that the Appli­cant had not thor­ough­ly read or under­stood the case law instead only read­ing ‘the excepts pre­sent­ed by the AI”. This result­ed in the Appli­cant bring­ing a case that should not have been brought cre­at­ing unnec­es­sary cost to the Tri­bunal and the Respon­dent.”

Some Obser­va­tions

While there may be a legit­i­mate role for the use of AI in employ­ment lit­i­ga­tion, giv­en the present state of devel­op­ment of AI appli­ca­tions, and the man­ner appli­cants are com­mon­ly using them, AI is caus­ing ongo­ing headaches for respon­dents, courts and the FWC. These issues are exac­er­bat­ed by appli­cants either not declar­ing the use of AI (as some­times required by pro­ce­dur­al rules) or not being entire­ly can­did in declar­ing the extent or nature of such use. 

There are two fac­tors that might lead to some san­i­ty being restored. 

The first is adverse costs orders being made against appli­cants, although the costs pro­vi­sions in the Fair Work Act 2009 (Cth) (FWA) do place some lim­i­ta­tions on the extent to which a court or the FWC can do this. Sure­ly, how­ev­er, using AI to pro­duce pur­port­ed evi­dence or sub­mis­sions that include unchecked hal­lu­ci­na­tions, and then being less than can­did about AI use when asked, is square­ly in the ter­ri­to­ry of an “unrea­son­able act” for the pur­pose of sec­tion 570 (for court pro­ceed­ings) or sec­tion 400A (for unfair dis­missal mat­ters in the FWC).

The sec­ond is a real­i­sa­tion by appli­cants that they are being had and the promise of AI being a sub­sti­tute for an expert advi­sor is large­ly a hol­low claim prop­a­gat­ed by hype mer­chants. While AI might, at this point, large­ly be an exam­ple of the “Emper­or’s New Clothes”, it is the unwit­ting employ­ee appli­cants rely­ing upon it being left exposed and vul­ner­a­ble in a court or the FWC, with only spe­cious doc­u­ments and false hope cov­er­ing their dignity.