Artificial intelligence is often presented as an invisible, frictionless technology. In reality, it has very real environmental, social and human costs – costs that are disproportionately borne by communities in the Global South.
This article builds on a simple proposal: AI tools such as ChatGPT should provide users with environmental impact feedback. It expands that idea to include a harder truth – that environmental impact cannot be separated from labour exploitation, data extraction, and unequal global power structures.
The Original Question: Should AI Show Its Environmental Impact?
The initial suggestion was straightforward: AI platforms could provide users with a simplified indication of the environmental impact of their usage – energy, water or carbon – either:
- As a simple scale (e.g. Minimal > High)
- Via weekly or monthly account summaries
- As optional notifications or dashboard insights
This would not overwhelm users with technical data, but would gently raise awareness in the same way energy labels or food nutrition scores do.
Technically, this is feasible. AI providers already track compute time, model usage and infrastructure load. The barrier is not technology – it is willingness and governance.
Why Environmental Impact Cannot Be Viewed in Isolation
Environmental transparency alone is insufficient if it ignores who pays the human cost of making AI work.
Behind “clean” interfaces sit vast global supply chains of:
- Data centres built with public finance in lower-income countries
- Workers exposed to traumatic content for poverty wages
- Communities losing control of their data, water and infrastructure
This is where the conversation must widen.
Global South Exploitation: The Hidden Cost of AI
Richard earned $2 per hour watching suicide videos, child abuse content and graphic violence for nine hours a day in Nairobi.
His work made global tech platforms “safe”. His US-based counterparts earned $20 per hour for the same work.
When Richard and 150 colleagues attempted to unionise in 2023, they were fired, blacklisted, and told they were merely “freelance contractors” with no rights.
This is not an isolated case – it is systemic.
Across Africa, Asia and Latin America:
- Public development finance is used to build digital infrastructure controlled by foreign corporations
- Local populations provide the data that trains AI systems valued at tens of billions
- Workers are paid poverty wages to absorb psychological harm
- Governments are pressured into trade agreements that prevent data sovereignty
Examples documented include:
- Biometric databases covering 1.2 billion people with foreign contractors holding full access
- Chinese-financed surveillance systems used against political opponents
- Only 45% of least developed countries having data protection laws, compared to 96% in Europe
Even physical infrastructure mirrors historical exploitation, with submarine cables following former slave trade routes – extracting data rather than people.
Assessing the Impact
| Area | Observed Impact |
|---|---|
| Environmental | High water and energy use concentrated in regions with weaker regulation |
| Labour | Low-paid, high-trauma work outsourced to the Global South |
| Economic | Value extraction without fair local return |
| Data sovereignty | Foreign control of sensitive national data |
| Democracy & rights | Surveillance infrastructure used against citizens |
Environmental impact reporting that ignores these dimensions risks becoming a form of greenwashing.
What Meaningful Action Would Look Like
For AI Companies
- Environmental impact summaries paired with labour and sourcing transparency
- Clear disclosure of where moderation and data work is performed and under what conditions
- Fair pay parity for equivalent work, regardless of geography
- Binding commitments on data sovereignty and local governance
For Governments and Funders
- Environmental and human rights conditions attached to development finance
- Mandatory local data protection laws and enforcement
- Support for worker organising and legal accountability
- Public ownership or co-governance of critical digital infrastructure
For Users
- Demand transparency – not just performance
- Support platforms and providers that publish ethical impact data
- Understand that “free” AI often hides unpaid or underpaid human labour
Raising Awareness: Practical Next Steps
If this issue matters to you, consider:
Surveys & Polls
- Run public surveys via Typeform or SurveyMonkey asking whether users want ethical and environmental reporting from AI tools
- Use LinkedIn polls to reach professionals in tech, ESG and policy
- Use Twitter/X or Mastodon polls to test public sentiment
Public Engagement
- Write blog posts, Medium articles or Substack newsletters
- Raise the issue in AI ethics forums and community discussions
- Engage civil society organisations working on digital rights and labour justice
Direct Pressure
- Submit feature requests and ethical concerns to AI providers
- Ask organisations you work with how their AI tools address environmental and labour impacts
Conclusion
Environmental impact indicators for AI are a good idea – but they must not become a distraction from deeper structural harm.
True responsibility means recognising that AI’s footprint is not only measured in kilowatt-hours or litres of water, but in human wellbeing, labour rights, and data sovereignty.
The resistance is growing. Workers are organising. Courts are listening. Communities are pushing back.
The question is whether AI companies – and their users – are prepared to do the same.