The Work Did Not Disappear: AI as a Reason for Dismissal and Alleged Union Discrimination at MNT
What Lies Behind the AI Justification? Automation, Subcontracting and the Reorganisation of Employment

On 18 November 2026, a small but revealing protest took place outside Istanbul Training and Research Hospital in Samatya. The Türkiye Health and Social Services Workers' Union, Türkiye Sağlık-İş, demanded the reinstatement of eight members dismissed by MNT Tıp, a subcontractor working at the hospital. One of them said she had been told that artificial intelligence was the reason for her dismissal. But, she added, their work had not disappeared: they were correcting incomplete and erroneous AI-generated drafts and completing the reports.
We follow that statement in this article. “AI has arrived, your job is over” is a sentence we will hear more often. We need to learn to read it from the standpoint of the working class.
Summary
- What happened? MNT provides subcontracted services to the Ministry of Health in the nuclear medicine departments of public hospitals. The union says eight of its members were dismissed because of union organising and has filed reinstatement lawsuits.
- Two reported explanations: A dismissed worker says she was given AI as the reason. The union says the employer also cited downsizing. We found no public statement from MNT, so these explanations are presented as reported by the worker and union, not as a direct statement from the company.
- Questions about the explanation: The union says three new employees were hired after the dismissals. Their roles and overall staffing levels must be examined before drawing conclusions about downsizing. According to Nisa Yılmaz, workers continued to correct AI-generated drafts.
- A separate allegation of union discrimination: The union describes a roughly four-year struggle for bargaining authority, two employer objections, and no disciplinary reports against the dismissed workers. In October, MNT workers had also publicly described alleged pressure and dismissals linked to union membership.
- Two distinct questions: Did AI eliminate the work? And who selected the workers for dismissal, using what criteria? The answers require evidence and must not be assumed in advance.
- Our argument: Automation often reorganises, divides, intensifies and cheapens work rather than simply eliminating it. Writing can become “correction”, and secure union jobs can be replaced with more precarious work. Neither outcome is a technological necessity.
- A reminder from the other side of the ocean: US unions have won contractual protections requiring advance notice of new technology, access to retraining or other positions, and severance in some cases. Screenwriters, performers, technicians and journalists have negotiated other safeguards. In Samatya, Yılmaz says she was given AI as a reason and was not paid her compensation; the public reports do not establish what advance notice workers received.
- Health matters too: Who corrects an erroneous draft, with what expertise and how much time, is a patient-safety question. The effects of dismissals and the arrangements for continued oversight need examination; no patient harm has been established in the reports cited here.
- Our position: We oppose the use of technology against workers, not technology itself. We demand reinstatement, an end to subcontracting, worker participation in decisions about AI at work, and the use of productivity gains to shorten working time.
Let us look more closely.
What happened?
We can reconstruct the events as far as public reporting allows, without filling the gaps ourselves.
October 2026. In a letter to Evrensel, MNT workers from the hospital's nuclear medicine clinic described alleged pressure after they joined the union. They said warnings not to join became threats of dismissal and that four workers were dismissed. They also raised serious radiation-safety allegations, including that cleaning workers had gone for years without dosimeters and that high-radiation treatments had been assigned to personnel they considered unauthorised. The workers criticised the company's award to the clinic's chief. Their account is an allegation requiring investigation; see also Evrensel's legal discussion.
18 November 2026. Türkiye Sağlık-İş held a press conference outside the hospital. The union said eight MNT workers were dismissed on various pretexts because they were members. Union representative Recep Kaplan said there were no disciplinary reports against them. Workers described colleagues with eight, twelve and fifteen years of service; another speaker, Dilan Güldü, said she had worked in healthcare for seventeen years (ANKA).
The AI explanation. Nisa Yılmaz, who had worked for MNT for two years and four months, said AI was cited when her contract was terminated. She said the work had not disappeared: workers corrected incomplete and erroneous AI drafts and completed reports. During busy periods, she said, she had worked until nine in the evening (ANKA).
Downsizing and three new employees. Kaplan disputed the reported downsizing justification. According to him, three new employees were hired at the same workplace after the eight dismissals (Başkent Postası; İşçi Haber). This alone does not establish the overall staffing change or whether the new employees performed the same duties.
The bargaining-authority dispute. Union branch chair Zerrin Ceren described a roughly four-year process, said MNT had twice objected to the union's authority, and said the union had secured recognition through the courts.
Code 22 and compensation. Yılmaz said her severance and other compensation were unpaid, that her social-security termination was recorded under code 22, “other reasons”, and that she could not receive unemployment benefit as a result. The dismissal-code list describes code 22 as other reasons. Entitlement to benefits and compensation depends on the circumstances and the relevant authorities' assessment.
Legal proceedings. The union says it filed reinstatement cases for all eight workers. It calls for reinstatement, payment of compensation, an end to public-sector subcontracting, permanent positions for technicians and secretaries, and an investigation by the authorities.
What is missing? The reports do not contain a response from MNT or the Ministry of Health. They do not establish the precise job titles of all those dismissed, which reports used AI, when the system was introduced, or how it was trained. The allegations and legal cases remain to be examined.
Two allegations, two separate questions
The reported reasons: Yılmaz says AI was cited to her; the union says downsizing was also invoked. In the absence of a public employer response, these are not presented as a direct company statement.
The workers' and union's account: The work remained, and the union alleges that workers were selected because they were members.
These issues must be considered separately. An employer could introduce a real technological change and use that change as an opportunity to dismiss union members. Technology may alter the work, but management decides who stays and who goes.
First question: Does the AI explanation withstand scrutiny?
- Is the work still being done? Yılmaz says workers corrected and completed drafts. Her account points to a change in the division of tasks, not the disappearance of all human work.
- Did total employment fall? The union says eight people were dismissed and three hired. That does not by itself establish whether the workforce shrank, but the new workers' duties matter when assessing the justification.
- How were people selected? The union says those dismissed were members. That raises questions about the selection criteria; it does not by itself legally establish discrimination.
Yılmaz's account challenges the simple claim that “AI finished the job”. At least the correction work she describes remained. What disappeared for the dismissed workers was their employment security.
Second question: The allegation of union discrimination
The union points to a long bargaining-authority dispute, objections by management, and the absence of disciplinary reports. Workers' earlier letter alleged pressure to resign from the union, requests for screenshots, transfers and threats. The union also reports new hiring. Yılmaz separately describes code 22 and unpaid compensation.
None of these is itself a court judgment. The courts will hear the cases. They are, however, reasons to investigate whether what is presented as technological necessity is also an organising dispute.
What does the law say?
This is a general outline, not legal advice:
- Under Labour Law No. 4857, an employer must show a valid reason when giving notice to terminate an indefinite contract covered by job-security provisions. Operational requirements can qualify; immediate termination for just cause is separately regulated. The employer bears the burden of proving a valid reason.
- In assessing operational dismissals, courts consider whether decisions are consistent and non-arbitrary and whether dismissal is a last resort. The duties of newly hired employees can matter in assessing a claimed reduction.
- Law No. 6356 provides for union compensation in cases of dismissal because of union membership, not less than one year's wages.
- Labour Law No. 4857 provides for joint responsibility in a qualifying principal-employer/subcontractor relationship. Whether that relationship exists and which claims it covers depend on the contracts and actual working arrangements.
The law on paper is not the whole question. Reinstatement cases can take months or years. In the meantime, a dismissed worker loses income, while others may fear what union membership could cost them.
Has the work disappeared, or is employment being reorganised?
A task is not a job
A job consists of many tasks. Preparing a health report may involve reading data, selecting findings, writing, checking, communicating with a physician and entering records. AI may accelerate some tasks, such as producing an initial draft. Checking, correction and professional responsibility remain. Finding an AI error can require more attention than producing text from scratch.
Yılmaz's account is an example of why the distinction matters. As we argued in The Human in the Loop, systems presented as autonomous often depend on human labour that becomes less visible when workers silently correct their outputs.
Marx on machinery
In Capital, Marx argued that machinery under capitalism is used to increase surplus value, not simply to make labour easier. By raising productivity it can reduce the working time necessary to reproduce wages and increase surplus labour.
He also observed that inventions could be used as weapons against workers' resistance and strikes. A machine is not only a production tool; it can also affect bargaining power. His criticism of the idea that displaced workers are automatically compensated by new jobs remains relevant: a future job elsewhere does not pay today's rent or guarantee the same security.
The problem is not machinery itself but its social use. Machines can shorten the working day and ease work; under capitalist relations they can also extend the day or intensify labour. That distinction grounds our position today.
Mental Taylorism
In Labor and Monopoly Capital, Harry Braverman described how management fragments work and appropriates workers' knowledge. Taylorist “scientific management” sought to transfer the craft worker's knowledge to management and reduce the worker to more easily replaceable tasks.
AI brings this process into forms of mental labour. We called it “mental Taylorism” in The Revolt of Crystallised Labour. A model trained on past reports could incorporate the products of experienced workers' knowledge; the data and training of MNT's system are not publicly established. A worker may then be told that the machine writes and the human “only checks”, before that checking is reassigned on worse terms. The union's account of eight departures and three hires makes these questions worth asking, not assumptions that can be treated as proven.
Five ways employment can be reorganised
- Fragmentation: A complete job is divided into machine and human tasks.
- Deskilling: Human checking can be dismissed as “mere correction” and paid less.
- Intensification: Workers may be required to check more drafts in the same time. Yılmaz described working until nine in busy periods; the reports do not establish that AI caused those hours.
- Precarisation: Experienced staff may be replaced by workers with weaker protections. The conditions of MNT's new hires need investigation.
- Disorganisation: A technological change can be used against union members. The union's allegation in this case remains a matter for the courts.
None is a necessary result of technology. Each depends on the social relations in which it is used.
Signals from elsewhere
- Radiology: Evrensel discusses continuing demand for radiologists despite AI's capabilities and the question of responsibility for decisions. In medicine, human oversight is not a luxury.
- China: A reported court decision concerned a worker who reviewed AI-generated content and was dismissed after refusing a lower-paid position. The court held that the possibility of using AI alone was not enough to justify the dismissal.
- Meta: Workers alleged in a lawsuit that an AI-generated dismissal list disadvantaged employees on protected leave and pregnant workers. The allegations require a judicial outcome.
- Meta's restructuring: Evrensel reported that a plan to transfer work to AI agents was shelved after employee resistance and disappointing results, following layoffs.
- South Korea: Hyundai workers protested the use of AI and humanoid robots in production, warning of effects on working conditions.
These examples show that AI can change the balance of power between labour and capital, not just the technical execution of a task. We have discussed related developments in Forty Percent of the Code, a Quarter of the Workforce and The Misery of the White-Collar Worker. The healthcare subcontracting system poses the same broad question in a different setting.
When workers write the rules: A reminder from across the ocean
Around the time MNT workers demanded reinstatement, Union Label reviewed safeguards US unions have negotiated against the risks of workplace AI. These do not resolve the Samatya case, but they show what workers can put on the bargaining table.
Las Vegas hotel and casino workers. Culinary Workers Union Local 226 and Bartenders Union Local 165 won contracts requiring advance notice before new technologies, including AI and robotics, are introduced. Affected workers can receive retraining or seek another position; if technology eliminates a job, workers can receive $2,000 in severance per year of service. A 2023 Caesars agreement provided six months' notice for introducing new technology and 30 days for changes in existing technology. Union leader Ted Pappageorge stressed that contract language must also cover technologies not yet known.
Screenwriters. The Writers Guild of America's 2023 agreement established that AI is not a writer, that AI-generated material cannot be used to classify a writer's assignment as a lower-paid rewrite, and that writers cannot be required to use AI. The 2026 agreement retained these protections and added notice when scripts or productions are licensed to train commercial generative AI systems.
This is instructive for MNT: the screenwriters sought to prevent “the machine produces a first draft, the human merely corrects it, and the correction is paid less” from becoming a way around their rights. Yılmaz's account raises a related question, though the occupations and contracts are different.
Technicians, performers and journalists. IATSE's agreement keeps covered AI-assigned work within the union contract and protects workers from being required to write prompts that displace another covered worker. Performers secured consent and compensation protections for digital replicas. Journalists at POLITICO and E&E News negotiated 60 days for bargaining before AI that materially affects their duties. After management introduced two tools without observing safeguards, the union filed a grievance; an arbitrator ruled for the union and management agreed to shut the tools down.
A common agenda: Advance notice, access to information, limits on surveillance and automated discipline, informed consent, fair pay and paid training. Grievance and arbitration procedures make these rights enforceable.
We should not exaggerate the gains. Notice is not a veto; severance does not stop a job disappearing. Axios notes that the great majority of US workers lack union coverage. Such provisions do not transcend capitalism, but they can constrain management's choices.
Compare that with Samatya: the union describes a four-year authority dispute and dismissals without disciplinary reports; Yılmaz describes an AI explanation, unpaid compensation and a code-22 termination that she says prevented unemployment benefits. We do not know whether the technologies used in the two workplaces are the same. The important difference is whether workers have the organised power to bargain over their use.
The slogan and the evidence
| “AI has arrived” | What is reported |
|---|---|
| The work was handed to the machine. | Yılmaz says workers continued to correct inaccurate drafts. |
| The business is downsizing. | The union says eight workers were dismissed and three hired; the roles and total staffing need examination. |
| This is a technical necessity. | The union says the dismissed workers were members and had no disciplinary reports. |
| Technology increases productivity. | Yılmaz says she worked until nine during busy periods. |
| Change is inevitable and nobody is responsible. | Yılmaz says her exit used the “other reasons” code and her compensation was unpaid. |
| AI improves services. | The account raises questions about continued human checking and patient safety. |
| This is only a company issue. | The service is provided in a public hospital. |
The right to health: Who corrects the faulty draft?
This is not only an employment question. Health reports affect patients. AI can assist in healthcare, but responsible use requires appropriate professional oversight. Evrensel's review discusses the reliability and limitations of AI in diagnosis.
Yılmaz's description of incomplete and incorrect drafts makes oversight especially important. If staff lack the experience, time or authority to check output, errors may go undetected. No patient harm in this case is established in the cited reports.
In February 2026, Evrensel reported concerns about outsourced MRI and tomography reports produced using copy-and-paste methods. Health workers argued for ending subcontracting. When cost cutting dominates a healthcare contract, AI can become a way to reduce costs rather than improve quality.
Subcontracting: The root of the problem
According to the union representative, more than 100,000 public-sector subcontracted workers are affected by similar issues. A worker provides a public service in a hospital but is employed through a company. This arrangement can produce:
- Diffuse responsibility: The hospital may point to the company, while the company points to the hospital.
- Lower costs and unequal rights: Workers performing similar tasks can receive different pay and protections.
- Obstacles to organising: A subcontractor may treat unionisation as a threat to its costs and business model.
AI gives management another possible explanation: “I did not dismiss you; technology did.” Public authorities remain responsible for overseeing services provided in public hospitals. They cannot simply hide behind a subcontractor. The specific legal relationship and liabilities must be determined from the facts.
Our class position
We reject two mistaken responses.
Hostility to technology: Calling for AI to be banned is an understandable reaction to insecurity, but breaking the machine does not change who owns it. Used properly and with oversight, AI can help healthcare workers. The question is who controls it and who benefits.
Technological fatalism: “This change is inevitable; adapt and improve yourself” presents a management decision as a natural event. The language of inevitability hides the people making that decision.
Our position is different:
- We oppose using technology against workers. If AI reduces labour, the gains should return to workers and society as shorter hours, less intensity and better services, not only company profits.
- Every AI justification must be scrutinised. What work disappeared? Did overall employment fall? How were people selected?
- AI at work must be subject to shared oversight. Workers and unions need information and a voice over the systems, purposes and changes involved.
- Healthcare AI requires public oversight. The system used, its error rates and arrangements for human review should be open to scrutiny, with appropriate professional responsibility.
- End subcontracting in permanent public services. The union's demand for secure positions for technicians and secretaries deserves support.
In Labour's Share we asked why labour's share can fall in the AI era. MNT shows why that question also belongs in a hospital corridor: workers are seeking reinstatement while Yılmaz describes her working hours and unpaid compensation.
Concrete tasks
For healthcare workers:
- Record when and for which tasks AI systems are introduced and how workloads change.
- Document errors without recording patient identities or personal health information, and follow legal and institutional data-security rules.
- Conduct a workplace survey on AI and workload together with your union.
For unions:
- Bargain over advance notice, joint assessment, job security, retraining and transfer rights, additional severance, protection against downgrading correction work, shorter hours from productivity gains, and limits on using worker data for dismissal or discipline. US unions have won different elements of this agenda in different sectors; Turkish workers can fight for protections too.
- Publicise and scrutinise dismissals attributed to AI rather than accepting the word as a sufficient explanation.
- Pursue reinstatement and union-compensation claims and investigate the responsibilities of the principal employer where the legal relationship applies.
For tech workers:
We also build, install and sell these systems. If a system becomes a reason to dismiss experienced healthcare workers, it concerns us too. Software workers also experience changes in how writing and reviewing are divided. As argued in Organising 2.0, solidarity networks across occupations are needed.
For readers:
- Share the MNT workers' demand for reinstatement.
- Ask about the conditions of the workers who welcome, care for and clean around you in public hospitals.
- Ask the health and labour ministries to investigate workplace AI and subcontractors in public hospitals.
Dear Young Comrades,
Think again about Nisa Yılmaz's words: the work had not disappeared. The report still needed completing, errors still needed checking, and patients still depended on human attention. The dispute concerns who does that work, under what conditions and with what security.
Technology is not a neutral wind. In this case the union alleges that a technological explanation was used to dismiss union members. Under collective control, the same broad possibilities could instead help shorten hours and improve services. Organised workers must have a say in the difference.
The machine is new; the chain is old. What can break that chain is something old and never outdated: standing together.
We will continue to follow the eight workers' struggle for reinstatement.
Sources
- ANKA: union seeks reinstatement of eight MNT workers; İşçi Haber; Başkent Postası: Recep Kaplan on downsizing and union membership.
- Evrensel: MNT workers' letter on radiation and organising; legal discussion.
- Evrensel: copy-and-paste MRI reports; AI and diagnosis; AI and unemployment.
- Evrensel: Meta's AI plan; South Korean workers and AI; GZT: China court decision; GZT: Meta lawsuit.
- Union Label: Workers Are Writing the Rules for AI; Bloomberg Law: Caesars agreement; Axios: union contracts and workplace AI.
- Social-security termination codes; Turkish Labour Law No. 4857; Law on Trade Unions and Collective Labour Agreements No. 6356.
- Karl Marx, Capital, Volume I, “Machinery and Large-Scale Industry”; Harry Braverman, Labor and Monopoly Capital.
Related articles: The Human in the Loop · Forty Percent of the Code, a Quarter of the Workforce · The Misery of the White-Collar Worker · Labour's Share · The Revolt of Crystallised Labour · Organising 2.0
This article is shared under the CC BY-SA 4.0 licence. It may be reproduced or adapted with attribution.









