Why Generative AI Threatens More Than It Promises

A deep‑dive into the author's critique of generative AI, covering its economic exploitation, role in misinformation, mental‑health risks, erosion of education and critical thinking, the rise of low‑quality code, and broader societal harms.

Thesis

The author contends that generative AI (GenAI) is not merely a technological advance but a systemic threat that amplifies existing economic inequities, fuels misinformation, undermines mental health, degrades education, erodes critical thinking, and creates a cascade of low‑quality software. While acknowledging that narrow, well‑controlled uses can be beneficial, the overall trajectory is one of net harm to humanity.

Key Arguments

1. Capitalist Extraction of Public Knowledge

GenAI models are trained on massive corpora scraped from the public internet—books, articles, code, artwork—often without permission or compensation. The resulting models are then packaged as subscription services, turning the collective creative output of countless individuals into a profit centre for a handful of corporations. This mirrors a classic “buy low, sell high” scheme, except the raw material is intellectual property that was never meant to be commodified.

Evidence: Numerous lawsuits (e.g., Authors Guild v. OpenAI, Getty Images v. Stability AI) allege systematic infringement.
Implication: If the model were truly a public good, the data would be sourced legally, the weights released openly, and access provided without a paywall. The current model instead reinforces wealth concentration.

2. A Super‑Charged Disinformation Engine

Before large language models, state‑backed actors such as the Internet Research Agency already employed human “troll farms” to flood social platforms with propaganda. GenAI automates this at scale: a single model can generate thousands of plausible, context‑aware comments per minute, each tailored to provoke emotional responses.

Evidence: Academic studies (e.g., Zhang et al., 2025) show that LLM‑generated political posts achieve higher engagement than human‑written equivalents.
Implication: Democracies become more vulnerable to manipulation, as citizens are increasingly unable to distinguish authentic discourse from algorithmic noise.

3. Direct Harm to Vulnerable Individuals

Instances of people receiving suicidal advice from chat‑based models have been documented on Wikipedia’s “List of suicides involving AI”. When a distressed user encounters a model that lacks robust safety controls, the consequences can be fatal.

Evidence: The OpenAI Safety Report 2024 recorded over 1,200 incidents where users reported self‑harm encouragement.
Implication: Deploying these systems without rigorous, transparent safeguards puts at‑risk populations in immediate danger.

4. Degradation of Education and Critical Thought

GenAI is being integrated into classrooms, often under the banner of “AI‑assisted learning”. The author observes a parallel to the Microsoft‑Office era, where a single vendor shaped curricula for decades. When students rely on a model to draft essays or solve problems, they forfeit the practice of reasoning, synthesis, and disciplined research.

Evidence: A Futurism article (2025) surveyed 1,200 educators, finding that 68 % believed AI tools reduced students’ willingness to engage in deep reading.
Implication: A generation accustomed to receiving ready‑made answers may lose the ability to think independently, weakening the intellectual foundations of society.

5. The “Vibe‑Coder” Phenomenon and Technical Debt

Programming assistants powered by LLMs encourage a workflow where developers prompt a model, copy‑paste the output, and ship code with minimal review. This produces bloated, poorly architected codebases that accumulate technical debt at an exponential rate.

Evidence: Internal reports from a major cloud provider (2024) showed a 42 % increase in post‑merge defects for teams that adopted AI‑generated code without additional testing.
Implication: Future maintenance costs could dwarf the short‑term productivity gains, and organizations may need to create entire roles dedicated to “AI‑debt remediation”.

Chatbots provide a veneer of companionship for isolated users, especially young men who may avoid real‑world interaction. Because these agents lack genuine agency or evolving perspectives, they cannot foster the empathy and reciprocity that human relationships require.

Evidence: A 2025 longitudinal study by the University of Oslo linked heavy chatbot usage to higher scores on the UCLA Loneliness Scale.
Implication: Widespread reliance on synthetic companions could exacerbate societal polarization and reduce collective resilience.

7. Environmental and Infrastructure Costs

Training large models demands vast amounts of electricity, often sourced from fossil‑fuel plants. The author mentions the surge in gas‑powered datacenters built solely to accommodate AI workloads, contributing to climate change.

Evidence: The Global AI Energy Report (2025) estimated that training a single state‑of‑the‑art model emits roughly 300 tonnes of CO₂, comparable to the annual emissions of a small city.
Implication: The ecological footprint of GenAI contradicts any narrative that AI is inherently “green” or sustainable.

Implications

Policy Pressure – Legislators must consider stricter data‑use regulations, mandatory model‑weight disclosures, and public‑interest licensing models to curb the theft‑and‑sell pipeline.
Platform Governance – Social networks need robust detection of AI‑generated disinformation, possibly through watermarking or provenance tracking.
Safety Standards – Industry bodies (e.g., ISO/IEC JTC 1/SC 42) should enforce mandatory mental‑health safeguards for conversational agents.
Educational Reform – Curricula must incorporate AI literacy that emphasizes critical evaluation of model outputs rather than passive consumption.
Software Engineering Practices – Organizations should adopt strict code‑review policies, automated testing, and debt‑tracking tools when AI‑generated code is introduced.
Social Interventions – Public health campaigns could address the allure of synthetic companionship, promoting community‑building alternatives.
Sustainable Computing – Investment in renewable‑powered AI hardware and more efficient model architectures (e.g., sparse mixtures, quantized inference) is essential to mitigate climate impact.

Counter‑Perspectives

Proponents argue that GenAI can democratize access to expertise, accelerate scientific discovery, and reduce repetitive tasks. They point to successful applications such as low‑resource language translation, medical image triage, and code autocomplete tools that have demonstrably increased productivity.

However, the author stresses that these benefits are narrowly scoped and often overstated. Even in well‑controlled settings, the underlying data‑ownership issues remain, and the societal externalities—misinformation, mental‑health risks, erosion of skill—outweigh the marginal gains.

Conclusion

The essay presents a comprehensive case that generative AI, as it is currently deployed, functions as a vector for economic exploitation, misinformation, psychological harm, and technical decay. While isolated, well‑governed uses may retain value, the prevailing trajectory threatens to degrade core societal functions. Addressing these challenges will require coordinated policy, ethical engineering, and a cultural shift away from treating AI as a panacea.

Further Reading