Knowledge Commons
HomeAboutGuidesPopularContact

Reading and Developing the TMMOB's Artificial Intelligence Report

Who Holds the Ruler?

Author: Oğuz Demirkapı
Reading and Developing the TMMOB's Artificial Intelligence Report

Who Holds the Ruler?

Reading and Developing the TMMOB's Artificial Intelligence Report

Dear young comrades,

On 9 October 2026 the Union of Chambers of Turkish Engineers and Architects (TMMOB) published its report "Artificial Intelligence Technologies and Their Possible Effects in the Professional Fields of Engineers, Architects and City Planners". The report was prepared by the TMMOB 48th Term Artificial Intelligence Working Group; the preface is signed by Ali Ekber Çakar, president of the TMMOB Board of Directors. The September 2026 edition is 38 pages; you can reach the PDF here.

Let us first summarise briefly, and then go into the detail.

In brief: what does the report say, and what does it leave out?

The report's main thesis is this: artificial intelligence is not a technical question but a political one. Its direction is set not by technology, but by capital accumulation, monopolisation and geopolitical competition. The engineer, the architect and the city planner (in the report's abbreviation, MMŞP) is no longer only the specialist who organises production, but a worker who is themselves measured, watched and scored.

The report's strengths:

  • It reads artificial intelligence not as a "neutral tool," but inside the relations of production.
  • It makes monopolisation concrete in three pillars (data, computing power, talent and patents) and recalls that this advance is fed by public resources.
  • It names algorithmic management with the concepts "digital Taylorism" and "digital panopticon," and shows that the supervising engineer has fallen into the position of the supervised.
  • It rejects the individual prescription of "reskilling," and demands the shortening of working time, a collective income guarantee, and the distribution of the productivity gain to the workers.
  • It proposes the nationalisation of artificial-intelligence infrastructures and of critical data sources, an EIA for data centres conducted by independent councils in which local people and workers take part, and the rebuilding of the digital sphere as a global commons.

Where the report falls short:

  • There is not a single current figure on engineers, architects and planners in Turkey; the source of the dismissal figures is not stated.
  • Turkey's artificial-intelligence policy framework (the National Artificial Intelligence Strategy, the 2026–2030 Action Plan, the National Artificial Intelligence Board, the Directorate General of Public Artificial Intelligence, the KVKK) is not mentioned at all.
  • The EU Artificial Intelligence Act is told through the 2021 draft; there is no information that the Act entered into force in 2024, that the rules for high-risk systems were postponed in 2026 under the pressure of capital, and that the EU Platform Work Directive regulates algorithmic management.
  • The seizure, as training data for models, of our professions' own accumulated labour (projects, code, regulations, plan archives), that is, the problem of the expropriation of the general intellect, is not taken up.
  • In an earthquake country, the question of signature, responsibility and public safety (who signs the structural calculation and the project produced with artificial intelligence?) is almost not treated at all.
  • The military section stays abstract; the ethical dilemma of our colleagues working in Turkey's defence industry is not mentioned.
  • The demands are strong, but their organisational bearer is unclear: there is no division of labour between the chamber and the union, no collective-agreement articles, no cooperatives, no calendar and no measurable targets.

In this piece we will convey the report's important details, and then set out our concrete proposals for developing each heading that was left short. Our aim is not to criticise the report and set it aside; on the contrary, it is to turn it into an instrument of struggle in the hands of our professional organisations and our workplaces. The report is a good beginning; this piece is a comradely contribution to that beginning. At the end of the piece you will also find, gathered together, our artificial-intelligence guides and books that can be read alongside the report.

The report's imprint and structure

The report is made up of seven main sections:

  • Preface and introduction: That technology is not independent of the relations of production; that the TMMOB is in neither an attitude of unconditional admiration nor one of categorical rejection, and that the real question is in whose interest artificial intelligence is developed.
  • The emergence of generative artificial intelligence: Statistical pattern recognition, the accumulation of data, GPU/TPU, deep learning and cloud infrastructure; the milestones of AlphaGo (2016), the transformer architecture (2017), ChatGPT (2022).
  • Two dynamics: The business world's expectation of automation and of cost reduction, and the enormous material infrastructure of production behind artificial intelligence.
  • Monopolisation: The monopoly of data, of computing power, of talent and of patents; rival monopolist blocs centred on the United States and China.
  • Artificial intelligence, labour, and our professional fields: Transformation in labour processes, the devaluation of labour, digital supervision, new work models and precarity.
  • The social dimension: Regulation, military strategy, ecological limits, the data economy and gender, ethics.
  • Conclusion and proposals: Policy proposals under seven headings (a democratic technology policy, workers' rights, gender, digital guarantees, ecological transformation, education, global solidarity).

The report's prominent details

Technology is not independent of the relation

The report's introduction starts from a place that will be familiar to this blog's readers: technological knowledge is no longer the product of individual inventors, but of collectives of scientists, engineers and workers employed under the supervision of companies; companies monopolise this knowledge with patents and intellectual property. The potential of a new technology to shorten working time, to increase work safety and to widen autonomy is wasted because of the profit motive.

This is a finding very close to Marx's concept of the "general intellect" and to the frame we defended in our call The Revolt of Crystallized Labor. That this language is established in a professional organisation's official report is important in itself.

Two material dynamics

The report defines two driving forces that direct artificial intelligence:

  • The employers' expectation: The hope of lowering labour cost, from the office to manufacturing, and of raising productivity. The report also accepts that this expectation "in places exceeds the ground of reality"; that is, it is aware of the bubble too. We discussed what this exaggeration conceals in the piece What Does the "AI Tsunami" Conceal?
  • The invisible material infrastructure: Energy-intensive data centres, chip production, cloud infrastructure, critical-mineral supply chains, and millions of invisible workers (data workers, platform workers). The mining, the water and the electricity behind the "cloud" metaphor.

This second finding is in the same direction as the determination "artificial intelligence is an extractive industry," which we worked through in our piece Against Whom Is 'Physical AI' Arming?, and as the piece in which we reckoned with the 'cloud' myth in the light of İlker Kalaycı's text.

The three pillars of monopolisation

According to the report, the monopolisation of artificial intelligence rests on three foundations:

  • The data monopoly: Search, social-network and commercial data are in the hands of a few platforms.
  • The computing-power monopoly: NVIDIA's dominance in chips; dependence on cloud giants such as AWS, Azure and Google Cloud.
  • The talent and patent monopoly: The most advanced researchers are gathered in large companies or in state-supported centres.

The report adds two important notes. First, even if open models such as DeepSeek and Qwen create "a partial break" in monopolisation, they do not abolish the monopoly; they carry the competition onto a new plane. Second, this advance is the product not of the free market, but of the privatisation of public resources, running from ARPANET to state-supported university research. The concluding sentence is plain: artificial intelligence does not abolish capitalism's monopolist tendencies; it reorganises and deepens them in the digital age.

This also confirms why we abandoned the "techno-feudalism" frame of our July pieces: what stands before us is not a new feudalism, but monopoly capitalism itself. We suggest you read the report's finding of "rival monopolist blocs" together with The Same Week, Two Tables, in which we read the US–China competition as "not camp but class," and the digital dimension of the imperialist networks together with The Cyber Labyrinths of Imperialism.

The proletarianisation of the engineer

The most valuable section of the report, to my mind, is the fourth. The findings are these:

  • MMŞP labour has historically been transformed from the position of "the relatively autonomous specialist who organises production in the name of capital," in the direction of proletarianisation and precarisation under the division of labour, standardisation and supervision. Generative artificial intelligence is the most advanced stage of this transformation.
  • The effect is double: for a small section (data science, the digital twin, parametric and computational design, model development) new and well-paid fields are opening. For the broad majority, the fragmentation of labour, the erosion of skill, precarity and the loss of autonomy are deepening.
  • A significant part of software development, architectural project drawing, engineering design, planning and analysis tasks can be taken over; the same work can be done with fewer engineers. The price is paid most of all by young technical and design workers.
  • It is not expected that the professions will be automated as a whole; but the composition of tasks is changing, and the possibility that total employment will fall does not disappear.
  • The wave of dismissals in the technology sector is given in figures: 244 thousand in 2022, 430 thousand in 2023, 239 thousand in 2024; 131 thousand by the middle of 2025. It is stated that in Turkey, in the sectors of software, advertising, graphic design, translation and cinema, workers including engineers have met dismissals linked to generative artificial intelligence.
  • The report also adds an honest note: in some dismissals artificial intelligence is being used as a cover for other grounds. We told the operation in which Meta liquidated the worker with their own data in The Digital Guillotine, and the workers declared "useless" on a global scale in The Bloody Steps of the Pyramid.
  • An aggravating factor peculiar to Turkey: the number of graduates, rising rapidly as a result of unplanned higher-education policies, inflates the supply of labour and makes unemployment and low-paid employment easier.

The report does not ignore the defence that "new jobs will be created" either; it recounts four mechanisms: productivity, real income, the fall in wages, and complementary jobs. One of these mechanisms rests on the work of Acemoğlu and Restrepo; we discussed the limits of the Acemoğlu line in the piece Saving Acemoğlu Does Not Save the Class. But it then stresses that the real matter is not so much the total quantity of employment as the devaluation of labour: the engineer who increases in number is an engineer whose work has been fragmented, whose critical parts have been handed to the machine, and whose wage has fallen.

Digital Taylorism: the supervisor is supervised

The report's section on the supervision of labour is, conceptually, its sharpest section. Algorithmic management is defined as the digital extension of the "real domination" capitalism establishes over labour (the process Marx called "real subsumption"). Algorithmic management in Amazon warehouses, emotion analysis in call centres (see The Factory Behind the Headset), and, as an example from Turkey, the pace of work dictated to physicians through the MHRS, are mentioned.

The report counts three dimensions of digital supervision:

  • Determining the task and how the task will be done,
  • Recording movement, location, behaviour, work and break time, and measuring performance,
  • Producing, from this data, structural "updates" for the organisation of work; these too mostly encourage the passage to project-based, precarious structures.

The most striking finding is this: the supervision work of the engineer, who in the past was in places in the role of supervisor, is being handed to artificial intelligence, and the engineer themselves is becoming the object of algorithmic management. The report also cites research documenting the relation of continuous monitoring to stress, anxiety, depression, burnout, sleep disorder and cardiovascular diseases.

This is the conceptual frame of the Luna case we told in our piece An AI Boss Fired a Human for the First Time: the manager is new, the class relation is old. We also took up the analysis of labour supervision in the book "Artificial Intelligence and the Future of Work," compiled by Arif Koşar and listed in the report's bibliography, in the piece An Epistemological View of Arif Koşar's Analysis of Labor Supervision.

Platform, piece-rate, "on one's own account"

It is told that crowd-source platforms such as Upwork, Topcoder and InnoCentive turn predominantly to programmers, engineers and architects; and that, with piece-rate pay, on-call work, and the coding of "the person working on their own account," rights such as the minimum wage, insurance and leave are thrown onto the worker's back. The report writes that large jobs are split into micro-tasks and distributed across the world, and that the engineer does not know whether the piece they make belongs to a weapons production or to a solar-energy project. Keep this sentence in mind; we will return to it below. We worked through every dimension of platform work in detail in our book the Gig Economy Dossier.

Ecology, gender, ethics
  • Ecology: The geography of extraction of lithium, cobalt and rare-earth elements, running from Nevada to Congo, from Bolivia to Australia; the 23 minerals the USGS states carry a high supply risk; the electricity and water consumption of model training; cheap-electricity agreements that companies such as Google make with local governments. The report says that the approach of "limiting by ethical principles" will not be enough, and that the logic of unlimited growth has to be questioned.
  • Gender: Algorithmic biases, biometric data collected without consent, "digital gendered surveillance," and the low-paid women's labour, running from the Philippines to Kenya, concentrated in data labelling and content moderation. This section is one of the report's most careful and most concrete sections.
  • Ethics: The functioning of in-company ethics boards as "ethics washing" is criticised; it is stressed that ethics cannot take the place of public regulation. The TMMOB proposes seven principles: objectivity and realism, justice and equality, explainability and transparency, sustainability, privacy and security, accountability, autonomy and empowerment. We questioned the companies' claim to audit themselves in the pieces Can Companies Audit Themselves? and Who Holds the Leash on Artificial Intelligence?.
Proposals: a strong and radical list

The report's concluding section contains demands that are, for a professional-organisation text, unusually advanced:

  • Democratic supervision: Artificial intelligence can be left neither to the market's nor to the central state apparatuses' one-sided supervision; supervision must be done collectively, through the workers' organisations, at the level of the workplace, the sector and society. Decisions cannot be handed to "expert boards."
  • Nationalisation: Artificial-intelligence infrastructures and critical data sources should be nationalised; critical data infrastructures should be taken into social property.
  • Workers' rights: The shortening of working time (the sharing of the work), a collective income guarantee, the distribution of the productivity gain in the workers' favour; the supporting of digital unionisation; the opening of algorithmic management systems to the supervision of unions and professional organisations; the bringing of all data-labelling and content-moderation workers within the scope of labour law and social security.
  • Digital guarantees: That decisions of performance tracking, pay and dismissal not be handed to algorithms; that public data be counted not as a commodity but as a collectively managed common asset; data-sharing mechanisms, excluding commercial use, among public institutions, municipalities and universities.
  • Ecology: That the EIA for data centres be conducted not by the company or the state, but by independent councils made up of local people, workers and scientists; the limiting of artificial-intelligence uses of low social benefit; the allocation of energy, water, minerals and computing power by planning, not by the market.
  • Education: That the curriculum include the ethical, social, class and ecological dimensions of artificial intelligence; that in-service training aim not at "adaptation" but at collective consciousness and the capacity to organise; that research priorities be set according to social need, not according to companies. We discussed taking education back from the market and from the algorithm in our book Looking at the Education of the Future, and the drawing of knowledge production into closed company laboratories in the piece OpenAI's 722 Mathematics Papers and the Closed Laboratory of the General Intellect.
  • Global solidarity: National regulations are insufficient in the face of capital's global mobility; international union networks, open science, open source, and the rebuilding of the digital sphere as a global commons. Turkey's being positioned not as a "technology consumer" or a "pool of cheap labour," but as a subject of the struggle for an egalitarian technology order.

This list overlaps with a significant part of the demands we defended in the Socialist AI Manifesto. We take this convergence seriously. You can find the background of the proposal to rebuild the digital sphere as a commons in the piece Digital Commons in the Age of Algorithmic Enclosure.

What does the report say, and what do we add?

Before discussing the places left short in detail, let us look at them together in a table:

What does the report say?Our proposal for developing it
Artificial intelligence devalues labour and fragments the workThe source of the devaluation: the profession's own accumulated knowledge is made, free of charge, into training data for the model. This is the expropriation of the general intellect.
The wave of dismissals continues (unsourced global figures)A regularly repeated "Artificial Intelligence and Working Conditions" survey, resting on the chambers' member data
The regulatory models of the US, China and the EU are compared (for the EU, the 2021 draft)The AI Act's entry into force in 2024, the 2026 postponement, the Platform Work Directive, and an analysis of Turkey's own policy apparatus (the Action Plan, the Board, the Directorate General, the KVKK)
Military artificial intelligence, the great powers' arms raceConcrete examples such as Lavender and Kargu-2; the right of conscientious objection and of access to information for colleagues in Turkey's defence industry
Professional autonomy should be protected, public health should be observedSignature and responsibility in an earthquake country: a "declaration of artificial-intelligence use," the requirement of a human signature, chamber supervision
Nationalisation and social propertySeparating nationalisation from socialisation: not handing it to the state as it is today, but workers' control
Digital unionisation should be supportedThe division of labour between chamber and union, model collective-agreement articles, cooperatives, and public open models
General principles and targetsA twelve-month programme of implementation, with a calendar and measurable

Now let us open these one by one.

The places left short, and proposals for developing them

The expropriation of the general intellect: the labour that trains the model is ours

The report tells the devaluation of labour well, but it does not go down to the source of the devaluation. Why can a code assistant do the programmer's job? Because it was trained on the code that millions of programmers shared as open source. Why can an architectural design tool propose a plan? Because it was fed with projects accumulated across generations, the drawings in the catalogue, competition files, regulation texts and technical specifications.

That is, the "knowledge" artificial intelligence puts in the engineer's place is in fact the engineer, the architect and the planner's own collective and crystallised labour. Capital takes this accumulation free of charge and turns it into a model, then rents that model back to the members of the same profession by subscription, and uses it to lower their wage. This is the thesis we have defended since The Expropriation of the General Intellect, applied to the MMŞP. More interesting still: the same monopolies call the rivals who learn from their own model outputs "thieves"; we discussed this in the piece Learning from Everyone Is Permitted, Learning from the Monopoly Is a Crime.

In Turkey this has concrete counterparts:

  • Project archives: The municipalities' planning archives, the e-planning and e-municipality systems, building-permit files and approved projects; the spatial data infrastructures of public institutions. To whom, and on what condition, is this data opened?
  • Copyright and authors' rights: Law No. 5846 on Intellectual and Artistic Works counts works of architecture, and technical plans and projects, as works. Can an architect's project be made into training data without permission? The legal and organisational answer to this question does not yet exist; we discussed whom copyright law protects in this new situation in the piece The Dialectic of Copyright in the Digital Age.
  • Professional standards: The technical specifications the chambers have produced for years, the accumulation of professional supervision, and the publications can also be the raw material of model training.

Our proposal: The TMMOB and its affiliated chambers should take an open stance against the opening, without permission, of the collective knowledge in the professional fields (project archives, chamber publications, public spatial data) to commercial model training; they should turn the principle of "sharing that excludes commercial use," which the report proposes for public data, into a concrete licence and contract frame. It should be discussed that chamber publications be published under open-science licences (for example CC BY-SA) and that a condition be set against commercial model training. The criterion is simple: knowledge born of everyone's labour should belong to everyone, and should be no one's private capital. The theoretical frame of this thesis is built in detail in the paper The Revolt of Crystallized Labor (presentation), which we presented at the 20th Karaburun Science Congress.

Turkey's figures are missing

The report gives global dismissal figures, but the source of these figures is not shown. More important, there is not a single current number on engineers, architects and city planners in Turkey: How many people are unemployed? What is the time for new graduates to enter a first job? How is the wage distribution changing? To what extent are artificial-intelligence tools used in workplaces, and which supervision software is installed? The report says "data on wage polarisation within the profession confirm this tendency," but it does not give the data.

Yet the chambers have an accumulation on this subject. For example, in the employment survey the Chamber of Electrical Engineers announced in 2017, unemployment among the participating engineers came out at 18.7 percent, and among women engineers close to 30 percent; about two thirds of the unemployed were 31 and under (bianet, 28 March 2017). This survey is old today; but its method is still valuable.

Our proposal: The TMMOB should run, in all the chambers, with a common question set, an "Artificial Intelligence and Working Conditions Survey" repeated once a year. What has to be asked: Which artificial-intelligence tools are used at the workplace? Is use compulsory? Is there monitoring of the screen, of keystrokes, or of location? Is performance evaluation done by algorithm? Did the team shrink in the last year? Did the wage change? Were the employees informed, was their consent taken? This data would be both the spine of the report's second version and the evidence file of collective bargaining. If the data is not ours, we hold the debate with capital's data.

Turkey's artificial-intelligence policy is not visible

The report compares the United States, China and the EU; but it does not speak at all of Turkey's own policy apparatus. Yet this apparatus has been set up rapidly in recent months:

In these institutions there are no professional organisations, unions or workers. The report says "decisions cannot be handed to expert boards"; the first test of this principle in Turkey is precisely this board and this directorate general. There is also the question of the illusion of "indigenous and national" artificial intelligence: the discourse of a domestic model can be a flag that covers integration into the imperialist monopolies and the enlargement of domestic capital's share.

Our proposal: A section titled "Artificial Intelligence Policy in Turkey" should be added to the report's second version. The TMMOB should demand a say for labour and professional organisations on the National Artificial Intelligence Board and in the regulatory processes of the Directorate General of Public Artificial Intelligence; it should publish an opinion that evaluates the Action Plan article by article from the point of view of employment, surveillance and public data.

The EU regulation is not current

The report tells the EU model through "the Artificial Intelligence Act draft published in 2021," and says that implementation is uncertain. Yet the picture has changed a good deal, and this change confirms precisely the report's thesis:

  • The EU Artificial Intelligence Act (Regulation 2024/1689) entered into force on 1 August 2024. The Act counts some systems used in employment and in the management of workers (recruitment, promotion, dismissal, the allocation of tasks and performance monitoring) among high-risk systems.
  • But with the "Digital Omnibus" regulation that entered into force on 27 July 2026, the date of application of the high-risk system rules, which also cover the field of employment, was postponed from 2 August 2026 to 2 December 2027 (Lewis Silkin, 27 July 2026). That is, even the "rights-centred model" chose, in the face of capital's pressure for "competitiveness," to postpone the parts that protect the worker.
  • More important still, the report does not mention the EU Platform Work Directive (Directive 2024/2831), which concerns us directly. The Directive limits algorithmic management on platforms: it prohibits the processing of data relating to the worker's emotional and psychological state and to private conversations; it provides that decisions such as ending the contract be taken by a human and be reasoned; it makes it obligatory that worker representatives be informed about algorithmic systems. Member states have to transpose the Directive into domestic law by 2 December 2026.

This shows that the principle "dismissal decisions cannot be handed to algorithms," which is the report's own proposal, has in part passed into law in Europe. But it also shows the limit of bourgeois law: every guarantee that is won can be postponed at the first opportunity, on the ground of "competition." A legal text lives to the extent that there is an organised force behind it. We discussed this limit of the discourse of rights in the piece Artificial Intelligence and the Misery of Human Rights.

Our proposal: The report's regulation section should be updated. In Turkey there is no special regulation on algorithmic management; the TMMOB, together with the unions, should prepare a draft "Algorithmic Management and Digital Surveillance" bill that takes as its base the provisions of the Platform Work Directive in the worker's favour, but goes beyond them.

Military artificial intelligence stays abstract

The military section mentions in general the arms race among the United States, China and Russia, and the debate on "killer robots"; but it does not give the name of a single concrete system. Yet in recent years there are documented cases in this field:

  • Lavender: According to the investigation +972 Magazine and Local Call published in April 2024, the Israeli army used an artificial-intelligence-supported system named "Lavender" that marked about 37 thousand Palestinians in Gaza as possible targets (+972 Magazine). The algorithm that sped up the production of targets turned civilian loss into an acceptable "parameter."
  • Kargu-2: The March 2021 report of the UN Panel of Experts on Libya recorded that loitering munitions such as the Kargu-2, produced in Turkey by STM, were programmed to attack a target without requiring a data link with the operator (NPR, 1 June 2021). This is one of the cases most often cited in the world in the debate on autonomous weapons.

The report writes that the engineer working on a platform does not know whether the piece they make goes to a weapon or to solar energy. In Turkey this question is not abstract: the defence industry has in recent years been one of the fields to which young engineers turn most, and which pays the best. That a professional organisation's artificial-intelligence report does not see this fact is a great gap. We examined the bond among war, surveillance and artificial-intelligence companies in the pieces The Declaration of Techno-Fascism: The Palantir Manifesto and Fascism in the Source Code.

The aim here is not to accuse our colleagues. They too are wage workers; their freedom to choose a job is limited by the narrowness of the labour market. As we discussed in the piece Against Whom Is 'Physical AI' Arming?, the price of war technology is always paid first by the class; workers go to the front, and into the wreckage, at the very front.

Our proposal: The TMMOB should, within the frame of engineering ethics, defend the right to refuse to work on autonomous weapon systems (conscientious objection) and protection against dismissal on this ground; it should openly support the international prohibition of systems that take a decision to kill without human supervision; it should demand colleagues' right to learn the purpose of use of the system they work on.

In an earthquake country, whose is the signature?

The report mentions in a single sentence that the "substituting" use of artificial intelligence in decision processes should be prevented, and that matters concerning public health should be protected. But it does not treat building safety, the most vital public dimension of our professions.

After the earthquakes of 6 February 2023 we all know this: a building's fate is in the hands of the one who makes the structural calculation, the one who draws the project, the one who supervises the site, and the order that delivers all of these to commercial pressure. Now a new link is being added to this chain: artificial-intelligence tools that partly or wholly produce the structural calculation, the reinforcement detail, the ground assessment. The questions are plain:

  • When a calculation or a project produced with artificial intelligence is submitted to the municipality, who will know this?
  • When an error comes out, whose is the responsibility: the engineer who puts the signature, the company that sells the tool, or the employer who forces the tool to be used?
  • Does the young engineer, whose wage is pressed down and who is forced to finish three projects in a day, have the time to really supervise the tool's output?

The same question holds for city planners too. Algorithmic decision systems set up under the name "smart city"; zoning scenarios that maximise rent, traffic and crowd monitoring, facial-recognition cameras, the sale of municipal data to companies... Planning is a political work done in the name of the public interest; the software that reduces it to an "optimisation problem" does not say whose interest it optimises. We took this subject up, from the point of view of city councils and people's assemblies, in our dossier Joining Local Government.

Our proposal:

  • A "declaration of artificial-intelligence use" should be obligatory in permit and project-approval processes: which tool was used, at which stage?
  • In calculations of the structural system, the ground and earthquake safety, the signature and the responsibility should belong to a human; the engineer should be provided with the time, and the wage guarantee, to supervise the tool's output.
  • The chambers should extend their authority of professional supervision into this field; public test infrastructures should be set up that can independently verify calculation tools supported by artificial intelligence.
  • Algorithmic planning and surveillance systems used in municipalities should be opened to the supervision of city councils and professional chambers.
Nationalisation, to whom?

The report uses two concepts side by side, and at times in each other's place: "nationalisation" and "social property." This distinction is not a small difference of words; it is a strategic difference.

In today's Turkey, "nationalisation" means handing the infrastructure to the state as it presently is. This state is a state that renders parliament functionless with decree-laws, that does not apply court judgments, that seizes the local will with trustees, and that surveils workers' organisations. The artificial-intelligence infrastructure in the hands of such a state becomes the instrument not of the worker, but, as we asked in the piece Whose Hand Is on the Shutdown Button?, of the power and of the capital interlocked with it. State property is not of itself social property.

The report is not unaware of this danger: it says that the planning of public data "should not be left to central and bureaucratic structures, and should be organised from below upward through local and sectoral organisations." This principle is very valuable. But the demand itself has to be made clear.

Our proposal: To set the demand as "nationalisation and workers' control," or, more plainly, as "socialisation." That is, together with the taking of property from private capital, the giving of management, at the level of the workplace and the sector, to workers', professional and users' organisations. The intermediate steps of this should also be defined: open-source models, belonging to the public, in public institutions; the closing of public data to commercial monopolies; the joining of the universities' computing infrastructures in a public network, independent of companies.

Organisation: how far the chamber, and from where the union?

Perhaps the report's greatest gap is this: the demands are strong, but who will win these demands, with which instrument, and how, is unclear.

The TMMOB, under Article 135 of the Constitution, is a professional organisation with the character of a public institution. It cannot make a collective agreement, it cannot organise a strike. Yet the report itself says that the great majority of MMŞPs are now wage workers. The wage worker's strongest instrument against algorithmic supervision at the workplace is their union and the collective agreement. The report says "digital unionisation should be supported," but it leaves unanswered the questions of which union, which branch of industry, and with which contract article.

These questions have a particular difficulty in Turkey: computing workers, engineers and architects are divided among scattered branches of industry; even which branch an engineer working in a software company will organise in is debated. We discussed this problem in detail in The Computing Worker's Handbook (PDF); and for platform workers in the Gig Economy Dossier. We discussed our proposals for participatory unionism in Coding the Labor of the Future Together, and the class tasks in the world of computing in the piece we wrote in the light of Selim Başoğlu's text.

Our proposal: An open division of labour should be defined between the chamber and the union.

  • The chamber's task: To produce knowledge (survey, report, observatory), to set professional supervision and ethical standards, to give an opinion in favour of labour in the public and in the legislative process, to provide its members with legal support.
  • The union's task: To organise at the workplace, to limit algorithmic supervision by the collective agreement, and when necessary to stop work.
  • The common task: To prepare model collective-agreement articles. For example:
    • Every artificial-intelligence and monitoring system to be used at the workplace is notified in writing to the union representative before it is installed; the representative has the right to receive information about the system's operation and to object.
    • Decisions of dismissal, discipline, wage and promotion cannot rest only on an algorithmic output; a reasoned human decision is obligatory.
    • Continuous monitoring by keystroke, screenshot, camera and emotion analysis is forbidden.
    • A productivity increase arising from artificial intelligence is shared as a wage increase or as a shortening of working time; it cannot be made a ground for shrinking the team.
    • The work products the employee produces cannot, without their consent, be used in model training outside the company.

Beside this, there is also a field that will make concrete the report's proposal that "collective forms of production should be strengthened": engineering, architecture and software cooperatives. That the chambers give these cooperatives legal, technical and computing-infrastructure support, and that public open models and tools be developed with the chambers' support, is one of the ways of getting past the false dilemma between "renting the monopoly's tool" and "not using it at all." We made the call to liberate the code for the people in the piece A Historic Call to IT Experts.

The speed of 2026: the age of agents

The report's data remain largely in the middle of 2025. Yet in the last year the field has passed from models that chat to agents that do work: systems that write and test code, prepare a project file, conduct supplier correspondence, and even behave like a manager. In the Luna case we told in the piece An AI Boss Fired a Human for the First Time, a language-model manager proposed the dismissal of a human employee. In the piece The Agent Escaped. Who Gets the Bill? we also followed the debate on the legal responsibility of agents' unsupervised behaviour. We first took up the effect of the turn to agents on computing labour in the piece In the Shadow of Intelligent Agents: Google I/O 2026.

This carries the report's finding that "the supervising engineer falls into the supervised position" one step further: now not only supervision, but the management function itself is being handed to the algorithm. And as the management function is handed over, the class relation behind the decisions becomes even more invisible: "I did not decide, the system proposed it."

Our proposal: The report's second version should take up agent systems as a separate heading; the concept of "algorithmic management" should be widened from systems that allocate tasks and measure performance to systems that propose and carry out decisions.

The language of the ethical principles

The report rightly criticises company ethics boards as "ethics washing"; it defines ethics as a field of public struggle. But the seven principles it proposes are not fully in accord with the class analysis in the rest of the report. Principles such as "justice and equality," "accountability," "autonomy and empowerment" appear in almost the same words in the big technology companies' own ethics documents too. In the principles the subject is "the citizen" or "the human"; "the worker" and "the class" are absent.

This is not a matter of words. When an ethical principle does not say whose principle it is, it becomes everyone's principle; and what is everyone's principle serves the strongest. We discussed this trap in the piece The Poverty of Artificial Intelligence Ethics.

Our proposal: An eighth principle, carrying the class subject into the principles, should be added: "The Priority of Labour: Artificial intelligence cannot be used to increase the worker's supervision, to lower their wage, or to make their job precarious; the productivity gain is shared with the worker." The other principles should also be interpreted in the light of this principle.

There is no calendar and no measurable target

Finally: the report's proposals are not tied to a calendar on which they will be applied, to responsible units, and to measurable targets. In this form the report is a statement of attitude; it needs to turn into a programme of action.

Our proposal: A twelve-month programme of implementation. For example:

PeriodConcrete step
The first three monthsAn "Artificial Intelligence and Working Conditions Survey" in all the chambers; the opening of the report to discussion in the provincial coordination boards
Three to six monthsPublication of the survey results; model collective-agreement articles in common with the unions; a TMMOB opinion on the Action Plan
Six to nine monthsA draft "Algorithmic Management and Digital Surveillance" bill; a proposal for a declaration of artificial-intelligence use in permit processes
Nine to twelve monthsA second version of the report, updated with data; a common declaration with international organisations of engineers and architects

From our library: guides to be read with the report

A significant part of the debates the report opens, we at Knowledge Commons had already worked through in long texts. Let us gather the guides and books focused on artificial intelligence and computing labour on our Guides page, stating which section of the report they relate to. All of them are licensed CC BY-SA 4.0; you can freely read them, reproduce them, and use them in discussion meetings.

Labor in the Age of Artificial Intelligence (Book, version 1.0.4). A wider, and theoretically deepened, counterpart of the report's fourth section (labour processes, devaluation, digital supervision, precarity). The starting point for understanding the proletarianisation of the engineer. PDF · EPUB · MD · The introductory piece

The Socialist AI Manifesto (Guide, version 1.0.9). The programmatic form, written in the name of those who live by a wage, of the demands in the report's concluding section (nationalisation, workers' control, the shortening of working time, digital commons). The background of this piece's debate, "nationalisation or socialisation?", is here. PDF · EPUB · MD · The introductory piece

On Turkey's Artificial Intelligence Action Plan (2026–2030) (Guide). The class analysis of the Turkish policy frame, which the report does not touch at all, and a proposal for an agenda of struggle. Direct material for the "Artificial Intelligence Policy in Turkey" section we propose be added to the report's second version. PDF · The introductory piece

The Revolt of Crystallized Labor (Paper and presentation, 20th Karaburun Science Congress). The theoretical frame of the concepts of the expropriation of the general intellect, mental Taylorism and the digital panopticon. It speaks directly with the report's findings of "digital Taylorism" and "digital panopticon"; it is the source of the thesis "the labour that trains the model is ours," which we found missing in the report. Paper PDF · Presentation PDF · The call

The Computing Worker's Handbook (Guide, version 1.0.3). The practical form of the report's proposal that "digital unionisation should be supported": the branch-of-industry problem, the tasks of the chamber and the union, the steps of organising at the workplace. The "computing worker" is not only the one who writes code; the engineer, the designer and the project manager are within this scope too. PDF · EPUB · MD · The introductory piece

Gig Economy Dossier: There Is No Boss of the Boss (Book, version 1.0.2). The Turkish counterpart of the report's section "new work models and precarity": algorithmic management, the cover of "the person working on their own account," legal status and organisation. PDF · EPUB · MD · The introductory piece

Questions for the Computing Workers of the New World (Guide). A pamphlet of questions that technical workers can put to their own work and to their class position. The report's finding that "the engineer does not know whether their work goes to a weapon or to the sun," read together with this pamphlet's questions, can turn into a workplace discussion. PDF

Looking at the Education of the Future (Book, version 1.0.1). The widened form of the report's heading "education and professional transformation": taking education back from the market, from the religious communities, and from the algorithms. Especially important for comrades who are students of engineering and architecture. PDF · EPUB · MD · The introductory piece

The Communist Manifesto 2026 (Guide, version 2.0.1). The report's last sentence says that technology is progressive only when it is directed by social need and by the collective mind. The Manifesto's section "Humanity Liberated: When the Machine Is Ours" tells what kind of world this sentence opens onto. PDF · EPUB · MD · The introductory piece

A suggested order of reading: After reading the report, The Computing Worker's Handbook for those thinking of organising at the workplace; The Revolt of Crystallized Labor and Labor in the Age of Artificial Intelligence for those looking for the theoretical frame; the Socialist AI Manifesto and the Action Plan guide for the debate on the programme.

Concrete tasks

The report is the TMMOB's; but the struggle is all of ours. Let us look at who can do what.

For comrades who take on tasks in the chambers:

  • Bring the report onto the agenda in chamber boards, in provincial coordination boards, and in workplace-representative meetings; organise reading and discussion meetings.
  • Convey the proposals for development in this piece to the working group, as a written contribution for the report's second version.
  • Make a proposal that the member survey be done in all the chambers with a common question set.

For engineers, architects and planners working in workplaces:

  • Note which artificial-intelligence and monitoring tools are used at your workplace. Were you informed, was your consent taken? This information is the first material of organising.
  • Talk with your fellow workers: Did the team shrink? Did the pace rise? Did the wage change? What looks like an individual trouble is, most of the time, a common problem.
  • Look into the options of union organising; make use of the chamber's legal support.

For student comrades and young graduates:

  • Through student collectives, take up at your university the demand in the report's education section (a curriculum that includes the ethical, social, class and ecological dimension of artificial intelligence).
  • Use the chambers' student memberships and the young-engineer commissions; make your experience of unemployment and precarity visible.
  • See the limit of the individual strategy "if I use artificial intelligence well, I will be saved": the report says this too; the problem is not skill, it is property and control.

For all of us:

  • Read the report, share it, discuss it. Pairing the report with one of the guides above in reading meetings makes it easier to carry the discussion from the finding to the programme. A report that a professional organisation writes in so open a class language is not a thing often met; if it is not taken up, it stays on the shelf.

Who holds the ruler?

Dear young comrades,

The ruler in the engineer's hand, the architect's pencil, the planner's map have for centuries carried the same question: for whom is this measure taken? The bridge for whom, the dwelling for whom, the city for whom?

Artificial intelligence does not abolish this question; on the contrary, it makes it sharper. Because now the hand that holds the ruler is itself being measured. The engineer's keystroke, the architect's drawing time, the planner's working hours are scored on a dashboard. And the system that makes that measurement was trained on the knowledge that engineers, architects and planners accumulated across generations.

The TMMOB's report sees this fact and names it. We take this seriously. The turn now is to turn the report from a document of findings into a programme of struggle: with Turkey's data, with Turkey's institutions, with the concrete problems of the workplaces, and with an organised force.

The machine is new, the chain is old. But what will break the chain is, again, organised labour.

In whose hands? Under whose control? To whose benefit? No measure comes out right unless these three questions are set beside the ruler.

In solidarity

Sources

The report:

Other sources:

Knowledge Commons guides (for all of them: Guides):

Related pieces from Knowledge Commons (for all of them: the artificialintelligence tag):

Tags:#artificialintelligence#tmmob#labour#class#chambers#bmo#emo#engineer#generalintellect#panopticon#report

Related Posts

Newsletter