In South Korea, the Korean Metal Workers’ Union (KMWU) is staging a historic strike against the use of artificial intelligence and automation systems in factories. Faced with the fear of being replaced by humanoid robots, workers began a series of phased walkouts from 15 July. The walkout, which began with two hours a day, will be extended to four hours from 20 July.

In January, Hyundai announced plans to use Atlas, a humanoid robot model developed by Boston Dynamics for industrial work, in its production. The union responded by stating that “even a robot using new technologies will not operate without an agreed contract”. The outlines of these plans were set out in May. During a meeting with investors, the company announced that more than 25,000 humanoid robots would be deployed across its production facilities.

Unionised workers, who are demanding job security from companies due to artificial intelligence, described their situation as follows: “News reports and videos showing that robots are becoming more advanced are causing workers to feel anxious about the future.”

HYUNDAI WORKERS DRAW THE LINE

Although Hyundai stated in its announcements that Atlas would be used as ‘muscle power’ and would only take on tasks that might be hazardous for workers, the union warned that this situation could lead to an ‘employment shock’ capable of abruptly altering working conditions across the entire sector.

The union had previously indicated it would go on strike over pay or the retirement age. Workers, who had reached an agreement through negotiations for some time, have now embarked on their first full strike since 2018. “Having seen the bonuses paid in the semiconductor manufacturing sector, we too are expecting better compensation this year,” said the union, drawing a line in the sand against the company.

Nearly 10,000 union members announced their strike plans at a press conference held in central Seoul. Workers, who wish to have a say in the integration of automation and artificial intelligence systems into production, are demanding one-third of Hyundai’s annual profits as a performance bonus, in addition to addressing their concerns about the future. This equates to a bonus of around 27,000 for 73,000 workers.

A similar call had been made by Samsung employees in recent months. Samsung, whose share price has risen by 300% due to demand for artificial intelligence and chips, did not change the bonus cap despite its employees’ performance. Following the threat of a strike, a settlement was reached to raise the bonus ceiling. Workers at Hyundai are also hoping for similar outcomes, whilst at the same time demanding assurances that their jobs will be safeguarded.

HOLLYWOOD WAS ALSO IN REVOLT

The ‘revolt’ against artificial intelligence is not limited to factories and production facilities; Hollywood has also risen up. In recent years, the Screen Actors Guild and the American Federation of Television and Radio Artists (SAG-AFTRA) have taken the fight against artificial intelligence – alongside the sector’s wider issues – onto the streets. During the strike, which lasted almost five months, screenwriters within the union put down their pens and actors walked off film sets. Following these actions, an agreement was reached within the film and television sector regarding protective measures against artificial intelligence.

THE THREAT OF INDUSTRIAL ROBOTS

Human-like robots, emerging alongside the advancement of artificial intelligence technologies, are set to replace workers. However, it is not only ‘humanoids’ that are driving change in the workplace. According to the International Federation of Robotics, the role of industrial robots in production has doubled over the last decade. Furthermore, algorithmic management systems—covering areas such as the allocation of tasks, monitoring and evaluation of staff—are also being integrated into workplaces. The training of these artificial intelligence models, which are reaching out to workers in various ways, falls once again to the workers themselves.

TRAINING THE ROBOTS THAT WILL TAKE THEIR PLACES

For textile workers in India, wearing a head-mounted camera has now become one of their daily tasks, alongside sewing and packing. Although managers at the factory have not explained the reason, workers are now obliged to record their work throughout their shift.

“At first, we thought it was funny,” said the textile workers, who, in an interview with The Guardian, described how the atmosphere had changed over time due to the pressure of being watched. They explained that they had begun to doubt their own productivity in their routine work and were acting more carefully to avoid making mistakes. What the workers, who are trying to carry on with their jobs under this surveillance, do not know is what these videos will be used for.

Recorded under the term ‘egocentric data’, these images show the workers’ hand movements and the production process. This data, collected by technology firms such as Objectways or Labellerr, will be sold to robotics companies to develop humanoid robots that could eventually replace workers. Textile workers are just the beginning. Millions of data points from workers in warehouses, shops and factories will also be required for comprehensive models.

EXPLOITATION CONTINUES AS EVER

India is a country that has gained a dominant position in this market. The concentration of labour-intensive industries opens up a wide range of opportunities for technology companies, whilst the hourly rate for data collection is less than half that of other countries. 60% of the data collected and processed here is sent to US firms.

However, the workers who produce this data have neither information about its intended use nor the right to object. In work environments where employment is precarious, workers are forced to consent out of fear of losing their jobs. In this situation, which opens the door to exploitation, whilst factories receive hourly payments from data collection firms, workers are in no position to even demand payment for the data they themselves produce. As technology companies take the competition in artificial intelligence to new heights, the labour of the working class left behind continues to fuel this sector.

WHAT THEY STEAL FROM THE ENVIRONMENT

Artificial intelligence does not merely harm workers and labourers; it also continues to deplete the environment and ruin the internet.

Globally, only 3% of groundwater is clean, and just 0.5% of all water is accessible and safe for human consumption. On average, a person can survive for only three days without water. Increasing droughts and water cuts are reducing access to water. Meanwhile, as data centres increase their energy consumption to meet the demands of artificial intelligence, they are consuming ever more water to cool processor chips and prevent overheating and resulting damage. According to experts at the University of California, every 100-word AI command is equivalent to one bottle of water. This may seem small, but when you consider the countless massive AI platforms such as ChatGPT and Gemini, the systems’ water consumption reaches alarming levels.

Every request you make to AI results in a request to a data centre; these data centres, which power the systems, consume an incredible amount of water to keep the systems cool. Even a medium-sized data centre alone consumes as much water as a small town, whilst larger ones consume as much as a city of around 50,000 people.

AI data centres are consuming far too many of the world’s resources. Artificial intelligence systems alone consumed 765 billion litres of drinking water last year – a figure exceeding the total consumption of the entire bottled water industry worldwide. Even if a portion of this water could be recovered (as little as 20 per cent), these water systems can still place a strain on areas suffering from drought or depleted groundwater resources. Google’s data centre in Arizona alone consumes 5.5 million cubic metres of water annually – enough to meet the annual basic water consumption of 753,000 people in need across Africa.

AI does not stop there; it also consumes a vast amount of electricity. If AI were a country in its own right, it would rank 39th in the world for electricity consumption in 2025 – placing it above countries such as Finland and Belgium. This electricity could meet the entire energy needs of Nigeria – the world’s sixth-largest country with a population of 224 million – for over two years. In 2023, US data centres were consuming 176 terawatts of electricity per hour, a figure roughly equivalent to that of Ireland, a country in its own right. Across the US, 56% of the electricity used by AI is generated from fossil fuels.

THE INTERNET IS NO LONGER HUMAN

The internet was originally designed on the assumption that there would be a human on the other side of the screen. This is no longer the case. According to the latest data released by Cloudflare’s Radar platform, AI accounts for more than half of internet traffic. According to the report, AI bots currently account for 57.4% of all global internet traffic, whilst humans account for just 42.6%. Cloudflare CEO Matthew Prince had predicted that robots would overtake humans by 2027, but this prediction materialised earlier than expected last June.

WHAT LED TO SUCH A RISE?

AI bots do not generate answers to the questions they are asked themselves; instead, they search for answers. Just as people do without AI, they visit various websites. As a result of people asking AI questions rather than conducting their own research, the majority of visitors to these websites are now AI bots. Although humans still lead when it comes to interacting with content, AI has overtaken humans in terms of website visits.

IS IT THE SAME IN EVERY REGION?

Of course, these figures vary from region to region. In North America, the bot rate is calculated at 68.6% and the human rate at 31.4%. In the American Midwest, however, the data shows a human rate of 54.5% and a bot rate of 45.5%. During peak internet usage hours, the bot rate in internet traffic in Gibraltar can reach as high as 97%. Other countries, such as Cuba and Laos, sit at the other end of the spectrum with human traffic ratios of 80.8% and 84.7% respectively. The data shows that whilst AI traffic is higher in larger regions, human traffic remains higher in smaller regions.

THE ‘DEAD INTERNET’ THEORY

Interest in the concept known as the ‘Dead Internet Theory’ has grown in recent years. The theory suggests that the proportion of online activities carried out by humans is decreasing. The main idea behind the ‘Dead Internet Theory’ is that the majority of internet activities are carried out by AI and bots. Whilst the theory was considered highly exaggerated when it first emerged in the late 2010s, it is becoming increasingly difficult to dispute this in light of Cloudflare’s reports. When additional findings are taken into account, the implications become even more concerning: 40% of Facebook content is generated by bots. In April, the music platform Deezer revealed that 44% of new tracks on its platform were created by AI. A report by Axios claims that 52% of all articles on the internet are generated by AI.

AI IS MOST PREVALENT IN LONG-FORM CONTENT

In a study by another platform, Pangram, the average AI ratio across all scanned content was 13.8%, although specific figures varied depending on the platform and the length of the content. On four out of five platforms, longer pieces of content were more likely to have been generated by AI than shorter ones. Across all platforms, a quarter of long-form content (25.72% of content exceeding 250 words) was generated entirely by AI.

Note: This article is translated from the original article titled Yapay zekaya karşı işçi isyanı, published in BirGün newspaper on July 17, 2026.