The Dark Side of AI: How Automation is Reshaping Working Conditions—And What We Must Do About It

During a recent webinar organized by the International Labour Organization (ILO) and the International Telecommunication Union (ITU), experts from around the world shared troubling accounts of how AI is already reshaping working conditions—from delivery couriers forced to chase algorithmically-determined targets to content moderators confronting graphic violence daily while training machine learning systems.

The central question, as ILO coordinator for digitalization and AI Sher Verick stated, is not whether AI will transform work—it already has. "The central issue is how to ensure that this transformation advances decent work and social justice," Verick said.

The Invisible Workforce Behind AI Systems

When most people interact with AI-powered platforms—asking ChatGPT a question, scrolling through algorithmically-curated social media feeds, or ordering food through a delivery app—they rarely consider the human labor that makes these systems possible.

Ben Richards of UNI Global Union described two primary groups comprising what he calls the "invisible" workforce powering artificial intelligence:

Content moderators who review harmful material to keep platforms safer, and data labelers and annotators who structure information so machines can learn from it.

"Wherever our organization speaks to such workers," Richards explained during the webinar, "they describe similar conditions: extreme pressure, constant monitoring, low wages and mental health harms."

These aren't isolated complaints from a handful of disgruntled employees. They represent systemic issues emerging across the global AI supply chain—issues that automation leaders and technology entrepreneurs must confront directly.

The Trauma Behind Your Screen

Major technology companies frequently outsource content moderation and data annotation work to countries in the Global South, where labor costs are lower and regulatory oversight is often minimal.

In India alone, tens of thousands of workers engage in this type of labor. For many rural residents, particularly women, job advertisements promising remote work requiring only an internet connection appear to offer rare opportunities for income and independence.

The reality proves far more disturbing.

According to recent media reports cited during the ILO-ITU webinar, workers often don't know what material they'll be expected to review until after they're hired. Many are required to sign non-disclosure agreements prohibiting them from discussing their work even with family members.

One woman from an Indian village described watching hundreds of videos daily, including graphic scenes of sexual violence, traffic accidents, and people dying. Another young woman reported being required to review content involving child sexual abuse and classify pornographic material.

These aren't edge cases. They represent the daily reality for a growing global workforce tasked with ensuring AI systems can distinguish harmful content from acceptable material—a critical function for platform safety, but one currently built on a foundation of worker exploitation and psychological trauma.

Human rights advocates have raised urgent concerns about these working conditions, yet the practice continues largely unchanged.

Algorithmic Management: The New Digital Overseer

Content moderation represents only one dimension of AI's impact on working conditions. Algorithmic management—where software systems determine work pace, task allocation, and performance evaluation—is creating new forms of workplace stress and safety risks across multiple industries.

The delivery and gig economy sectors offer particularly stark examples.

A 2025 study published by the University of Cambridge found that approximately two-thirds of drivers and couriers in the United Kingdom work under conditions of anxiety due to "unfair feedback" and sudden changes in working hours determined by algorithms. More than half of respondents indicated they risk their health and safety at work to meet system-imposed demands.

Trade union monitoring has documented fatal accidents linked to couriers "chasing impossible delivery targets set by algorithms," according to Evelyn Astor, director of economic and social policy at the International Trade Union Confederation (ITUC).

Critically, platforms typically don't explicitly instruct workers to violate safety rules. Instead, the system of incentives—including penalties for slow performance, speed-based bonuses, and priority order allocation—creates conditions where workers feel compelled to make dangerous decisions to preserve their income.

The algorithm doesn't tell you to run a red light. It just makes it economically devastating if you don't.

Beyond Delivery: AI Management Across Sectors

The concerns extending beyond the gig economy are proliferating across industries. Automated systems now assign shifts, set pay levels, and even make termination decisions in warehouses, call centers, retail operations, and professional services—often with minimal human oversight and limited avenues for appeal.

Trade union representatives at the webinar warned that deploying AI without proper safeguards risks reinforcing existing workplace problems while creating new ones:

• Work intensification as algorithms optimize for maximum productivity without accounting for human limitations
• Increased digital surveillance that monitors workers' every movement and keystroke
• Biased decision-making when AI systems trained on historical data perpetuate past discrimination
• Reduced human oversight in employment decisions affecting workers' livelihoods and dignity

Without appropriate regulation, Astor warned, artificial intelligence could deepen existing risks rather than solve them.

A Personal Perspective: Automation with Human Dignity

As someone who has built a career advocating for intelligent automation and teaching others to harness its power, these findings demand serious reflection.

"The automation revolution I've dedicated my career to building must prioritize human dignity alongside efficiency," I stated in response to these reports. "Technology that optimizes productivity while destroying workers' mental health or physical safety isn't innovation—it's exploitation with a digital interface. Those of us leading the automation movement have a responsibility to ensure the systems we build augment human capacity rather than diminish human worth."

This isn't anti-technology rhetoric. It's a recognition that automation's benefits depend entirely on how we choose to implement it.

At the Automation Institute™, we train thousands of students to develop efficient workflows through AI technology. But efficiency divorced from ethics is ultimately unsustainable. Systems that burn through human workers like disposable resources will eventually face regulatory backlash, public rejection, and workforce collapse.

The question isn't whether to automate—it's how to automate responsibly.

Making AI "Trustworthy": Standards and Governance

Bilel Jamoussi, Deputy to the Director of the ITU Telecommunication Standardisation Bureau, emphasized that AI is now being used in systems "with real consequences for people's prosperity," including hiring decisions and access to essential services.

Technical standards can help make AI "trustworthy," Jamoussi noted, but the decisive factor remains how societies and governments choose to apply these technologies.

This distinction matters enormously. We can develop the most sophisticated AI governance frameworks imaginable, but without political will to enforce them and corporate commitment to implement them, they remain theoretical exercises.

The ILO and ITU are advancing initiatives aimed at shaping international approaches to regulating AI's impact on labor markets, including the AI for Good platform and the Global Coalition for Social Justice.

At the first meeting of the Independent International Scientific Panel on AI held recently, UN Secretary-General António Guterres emphasized the body's "huge responsibility" in "helping shape the trajectory of artificial intelligence for the benefit of humanity."

"Collectively, you represent something the world has never seen before," Guterres told the panel members, who bring diverse expertise from multiple regions and disciplines.

Building a Global Alliance for Worker Protection

UNI Global Union is building a global alliance of content moderators and promoting safe-work protocols grounded in the right to organize and engage in collective bargaining.

"We want AI to augment human capacities," Richards stated. The benefits of technological progress must be distributed fairly, he stressed—not concentrated among platform owners while risks and harms are offloaded onto vulnerable workers.

This approach recognizes what technology evangelists sometimes forget: automation doesn't exist in a vacuum. It operates within social, economic, and political systems that determine who benefits and who bears the costs.

The most sophisticated AI in the world cannot compensate for governance structures that permit exploitation.

What Automation Leaders Must Do Now

For those of us building automation systems, consulting with companies on AI implementation, or training the next generation of automation specialists, these findings present both a challenge and an opportunity.

The challenge: Acknowledge that the automation revolution we're championing has real costs being disproportionately borne by the most vulnerable workers in the global economy.

The opportunity: Lead the movement toward responsible automation that genuinely augments human capacity rather than extracting value while externalizing harm.

Practically, this means:

1. Transparency in AI deployment
Companies implementing algorithmic management systems should clearly communicate how these systems work, what data they collect, and how decisions are made. Workers deserve to understand the systems governing their livelihoods.

2. Human oversight and appeal mechanisms
Automated systems should augment—not replace—human judgment in consequential decisions. Workers must have clear pathways to appeal algorithmic decisions affecting their employment, compensation, or working conditions.

3. Fair compensation for AI training labor
Content moderators, data annotators, and other workers in the AI supply chain deserve compensation that reflects the psychological burden and skill required for their work, along with mental health support and reasonable working conditions.

4. Safety-first system design
Algorithmic management systems should be designed with worker safety as a primary constraint, not an afterthought. Incentive structures that encourage dangerous behavior to meet performance targets represent design failures, not worker deficiencies.

5. Worker voice in AI governance
Those most affected by AI systems should have meaningful input into how these systems are designed, deployed, and modified. This includes both formal collective bargaining where possible and participatory design processes that incorporate worker feedback.

The Path Forward: Technology Serving Humanity

The United Nations system identifies the central challenge as ensuring that artificial intelligence expands human potential rather than undermines workers' safety and wellbeing. This requires shifting focus from technological innovation alone to governance grounded in human rights, equality, and sustainable development.

For the automation industry, this moment demands honest reflection about the gap between our sector's rhetoric and reality.

We speak enthusiastically about AI augmenting human intelligence, freeing workers from repetitive tasks, and unlocking new forms of creativity and productivity. These benefits are real—I've witnessed them firsthand in the automation systems we've built and the workflows we've optimized.

But we cannot celebrate these wins while ignoring the content moderator reviewing child abuse imagery for poverty wages, the delivery courier risking injury to meet algorithmic targets, or the warehouse worker whose every movement is monitored and optimized by systems that treat humans as inefficient components to be upgraded or replaced.

The automation revolution's success will not be measured by productivity gains or market capitalization. It will be measured by whether we managed to harness these powerful technologies in service of human dignity and broadly shared prosperity.

Those of us with platforms, influence, and expertise in this space have a responsibility to shape that outcome—not through empty gestures, but through concrete commitments to transparent, accountable, and genuinely human-centered automation.

The technology will advance regardless. The question is whether we'll ensure it advances humanity along with it.