Can we use AI and still be ethical?

Ethical considerations we should all be aware of when using AI.

Photo by Art Institute of Chicago on Unsplash

The narrative sells the idea that AI will bring prosperity to all, but is powerful enough to bring chaos and destruction. A handful of AI frontier labs and their visionary CEOs are the only ones equipped to steer the technology towards a positive pathway.

Sounds like a compelling story? Or one far too convenient for companies who are under huge pressure to justify the massive investments in infrastructure, and must maximize their company valuations for going public (IPO)?

Additionally, this narrative spreads ‘hype’ or ‘fear,’ which are central components for controlling the public debate, subconsciously manipulating mass behavior, numbing our collective critical thinking, and facilitating a power grab.

Is the future really going to be written according to those narratives? Or can we shape a different one for the future?

There is another pathway

Photo by Patrick Perkins on Unsplash

Big AI techs and their Messianic CEOs are pushing on us a vision of the future that may not include many of us, may not agree with, and may not trust to be a good one. As Mira Murati shared at her last interview when asked about her trust in her former boss, Sam Altman, she replied that this is less about the character of any individual leader (although this also matters), but more about the absence of structural checks: “Too much attention has been paid to virtue and too little to governance, she suggested.”

Photo by Markus Spiske on Unsplash

But I believe we can help define an alternative pathway by having better understanding of AI ethical implications, how each player performs ethically, and how we can make conscious use of AI, including the choices we make when selecting the AI company, model or assistant to be used. But apart from the well known big AI labs, there are several initiatives in the AI space proposing an alternative pathway to AI. One that advocates for a more balanced power distribution and can help us reclaim the control that most people are delegating to a handful of companies without even knowing. Those initiatives range from independent and investigative media, industry watchdogs, universities, scientific institutions, alternative open source LLMs, AI open source communities and more.

Key ethical considerations for a conscious AI use

We must all adapt to a new reality and AI is part of it, but how can we use AI without feeling we are disregarding all ethical implications of such technology? One thing at our reach, is to make a conscious use of AI. But exactly that means?

The conscious use of AI involves leveraging artificial intelligence to produce efficiency gains while minimizing impacts on environmental sustainability, ethical risks and hidden costs. This approach requires purposeful selection of AI tools, transparent data handling, and critical verification of outputs rather than simply accepting what AI generates.

#1. Data privacy and training practices:

Photo by Tasha Kostyuk on Unsplash

Several AI companies use your data — including prompts and conversations — to train their models. Most platforms have opt-out settings that are difficult to find, and many are changing their privacy policy to allow them to retain your data.

• Amazon decided to use voice captures from Alexa devices to train its models. Additionally Amazon faces lawsuit for capturing facial recognition data from other devices they sell, such as the video door bell Ring.
• OpenAI retains your data for up to 30 days after a deletion request, and basically collects all your data, including personal information, location, engagement, usage patterns, all content you input or generate (prompts, files shared, images, etc) and metadata about your device, browser, connection, device you are using, and who in your contact list also uses their service.
• Google also collects a lot of your data for training purposes.
• Claude used to use your data for training purposes only if you allowed it. However, recently, with the release of its new model Fable, the company changed its data retention policy and now collects and retains your data without exception.
• Perplexity AI data usage for training is enabled by default, but you can opt out by turning off the AI data retention toggle in your Account Settings under the Preferences tab.

Beyond the model training, your data can also be used to improve and develop company's services, research of new features; to personalize and customize user experience across services, and provide with more relevant content, which raise questions around how you can be influenced or manipulated by consuming generated knowledge with intentional or unintentional biases. Who knows?

#2. Bias and safety governance:

Photo by Storyzangu Hub on Unsplash

A biased AI system will make decisions based on stereotypes and existing societal biases such as racism, sexism, and other forms of discrimination, with the potential to impact millions of people. It has the power to amplify oppression and favor dominant ideologies, which can exclude the vulnerable people, and suppress diversity, fuel narratives against certain groups and suppress dissident voices. Additionally, it is hard to separate what is "hallucination" from bias, which makes it all more complex. Big AI labs have teams, called red teams, who work to reduce bias, but no external audit was ever made, making the reliability of it's internal of its mechanisms highly questionable.

#3. Labour & supply chain ethics:

Photo by Polly Sadler on Unsplash

Here is where most big AI labs fail. Behind most of the "intelligent" models there is an army of poorly paid annotators, mostly in poor countries, working in slave-like conditions and long hours to tag, classify, and filter data/content that is at best psychologically disturbing. Workers in Africa report long shifts that can go up to 20 hours a day, with exposure to graphic violence and abuse under no meaningful regulatory protection.

#4. Militar applications and Surveillance:

Photo by Mohammed Ibrahim on Unsplash

The lack of transparency and regulation is even more disturbing when it comes to AI Military application where the stakes are life and death as we cabn picture in the real story below:

Shajareh Tayyebeh elementary school in Minab, in southern Iran, tries to have one more day of class in the middle of the war. Its building had been separated from close to an adjacent Revolutionary Guard base separated by a fence and repurposed for civilian use nearly a decade ago. Apparently this fact was never updated. The result was the deaths of 168 people killed by an error, most of them children, girls aged seven to 12.

In Gaza other cases were reported and show how AI for military purposes is the most urgent ethical considerations related to AI use. AI intelligence can power mass surveillance, and technologies such as facial recognition allow AI to track dissident voices, and label marginalized populations as targets. Autonomous weapon systems hand the decision to kill to AI algorithm which will inevitably fail due to lack of context, poor judgment, or bias. As highligthed by Jeroudi Leith (2025):

If an autonomous platform operates erroneously, culpability grows unclear, creating uncertainties about both legal responsibility and ethical justification.

For example, a soldiers hesitation to kill is mechanism that is completely removed, and this completely dehumanizes people in the war zone, as the deaths caused by mistake are accepted as part of model's probability error rates. A school in Nuseirat (Gaza) was struck on July 7, 2024, where 23 people were killed, and according to Israel Defense Forces (IDF) only 8 were member of Hamas. How about the other 17 people killed? As reported from sources within the IDF:

… operators referred to low-level targets as “garbage” and accepted that most casualties were women and children. The system also allowed for a “collateral damage threshold” of up to 20 civilians per strike, applied automatically without assessing the actual threat posed by each target."

Photo by Aoun Abbas on Unsplash

For instance, Gould (2025) argue that AI proponents defend that more precise targets would reduces civilian deaths, but in practice AI accelerates the kill chain, from identifying the target to launching and attack. In her word citation:

“For example, during recent conflicts, Israel identified significantly more targets thanks to AI systems. Before the Gaza War, they might identify around 50 targets a year. During the war, that number increased drastically to as many as 100 targets per day. And Israeli officers had just 20 seconds to verify the AI-generated information and decide whether a target was legitimate.”

Photo by Diana khwaelid on Unsplash

According to the Lawfare Institute, the political moment and financial pressure are changing corporate culture and making them more tolerant to get involved in Defense work. For example, OpenAI in January 2024 revised its usage guidelines to lift restrictions that had explicitly barred its technology for applications linked to “weapons development” and “warfare.” Google also dropped its internal restrictions for the use of its AI on weapon systems. Something quite ironic for a company that in its origin used to consider its slogan: "Don't be evil". These movements reveal how big AI labs are playing with ethical boundaries and exploring the lack of global governance guidelines to create new forms of power and corporate nonstate sovereignty.

Photo by Igor Omilaev on Unsplash

The article "Military AI Policy by Contract: The Limits of Procurement as Governance" describe the corrupt dynamics being led by US government and major AI labs to skip the regular legal proceedings for the adequate and regulated procurement process and producing a contract deals that are intentionally flawed in its content. Unlike statutes, those contracts do not provide democratic accountability, nor public deliberation, and were designed with guidelines for setting up vague terms in critical clauses making them open to loose interpretation, and likely to facilitate abuse of power, and reduced public oversight. Guardrails are not even enforced by contract, but rather by vendor's technical controls.

#5. Environmental Impact:

Photo by Hermes Rivera on Unsplash

At every prompt we do, energy is being consumed, water is being consumed, carbon is being emitted in the atmosphere. US and Chine are projected to be the biggest natural resource AI consumers, and both rely on fossil fuel for energy production.

Massive AI data centers consumes a huge amount of energy and non-renewable resources to be constructed and run. Several studies predict that by 2030, AI electricity consumption could hit 1,500 TWh. To put this into perspective, this is the current total electricity consumption of India. Besides huge amounts of energy, AI data centers consume a vast amount of water, and risk to put even more pressure into a natural system that is already stretched.

The environmental risks include acceleration of carbon emissions and climate change, water shortage or price hikes for water and energy in regions where data centers are being built, let alone several indirect impacts along the supply chain (ex: material extraction, construction, land degradation, etc).

[embed]The illusion of AI efficiency AI promise of unprecedented efficiency may not be the full storymedium.com

What now?

Photo by Markus Spiske on Unsplash

Does everything seems doomed? Don't panic and take action! Below are few things you can start exploring, and at the end a list of AI choices that may help you stay true to your values, and which companies should be avoided given their low or declining ethical standards.

Protecting your data privacy

Prioritize open-source lightweight models, models that can run locally on your hardware, for less complex tasks. This can reduce your AI costs, as well as can add protection to your data privacy. Select providers that never transmit your data, or AI wrappers that provide a constitutional guarantee of no training.

Avoiding model's biases

In order to avoid model bias, use prompts that explicitly instruct the model to critique potential bias and propaganda, and to use a diverse set of mainstream and independent media sources, as well as respected and recognized institutions for research. Favor companies with published constitutions (ex: Thaura.AI), responsible scaling policies, and with values transparently disclosed. Ensure human review to avoid discriminatory bias in high stakes decisions (hiring, financing, medical access).

Standing for non-abusive labour practices

Where possible, avoid AI companies with documented labour abuses. OpenAI, Meta, Google are some examples of companies that hire AI training providers that profit by exploiting vulnerable people in poor countries. If possible, choose local AI tools (which minimize moderation needs) or open-source ecosystems (ex: Hugging Face) with community governance. Support initiatives that can help reduce abusive labour practices such as the Data Labelers Association, or by demanding more transparency from tech companies.

Rejecting AI uses for militar and surveillance purposes

Avoid major defense contractors and companies with documented surveillance ties. Prioritize tools that reject militar use for surveillance or killing systems or explicitly bans the militar use. Support campaigns such as "No Tech for Apartheid" organized by Google and Amazon employees to halt Google’s Project Nimbus, and "Stop Killer Robots" that advocate for global international laws to regulate AI military uses and applications.

Reducing your environmental impact

For light personal use (<1 hour/day), local AI on efficient hardware (ex: Apple M-series chips) can reduce energy consumed in transmission and data center overhead. For heavy use or teams, explore ethical cloud wrappers hosting based on renewable energy that can be provide better performance per query than running local GPUs 24/7.

Be selective with your AI use as not all tasks or flows are worth automating. For example, many tasks may not require an advance model or a heavy processing. Unless you a software engineer working in a large and complex codebase, scientist doing advanced research across multiple sources, or a security specialist running a complex penetration test; you may not need to use an advanced model. Open source models can be used with high success rates for tasks such as summarizing a document, classifying an email into categories, extracting structured data from a form, translating text, generating a first draft from a brief, routing a customer support ticket.

Advocating for more transparency on model training

Another reason for advocating for Open Source models is related to the transparency on what researches and scientific data the model is been trained on. As Spirling (2023) points:

"The most widely touted LLMs are proprietary and closed: run by companies that do not disclose their underlying model for independent inspection or verification, so researchers and the public don’t know on which documents the model has been trained. The rush to involve such artificial-intelligence (AI) models in research is a problem. Their use threatens hard-won progress on research ethics and the reproducibility of results."

Additionally, you can support EU AI Act requirements for energy disclosure (mandatory post-2027) and demand companies to be transparent about their product's lifecycle (ex: manufacturing, training, inference, disposal).

Stay true to your values

Getting informed about AI companies that forfeit ethics in favor of profit or investment is the first step in practicing the conscious use of AI.

Photo by The Cleveland Museum of Art on Unsplash

If you share the same ethical concerns described in this article, be aware about companies who amy incur into some of these practices:

• Mass surveillance: Amazon,
• Abusive labor practices for model training: Open AI, Meta, and Microsoft
• Developmend of autonomous weapons systems: Palantir, xAI (Grok), c3 AI, and G24 AI. OpenAI and Google are also reportedly involved in contracts that allow the US Army to use their models for any security related purpose.

Photo by Greg Rakozy on Unsplash

There are several AI initiatives out there worth knowing (and perhaps supporting) that have the potential to tackle the main ethical concerns:

• Models that can run locally: LM Studio, Ollama, Jan AI, Github, Deepseek, and Nomic
• AI wrappers or search wrappers that prioritize data privacy such as Thaura AI, Duck AI, Okara AI, and Lumo (Proton).
• Companies who are based on renewable energy sources such as GreenPT; Thaura AI

By doing so, you might be able to use AI while still remaining true to your values and influence the AI market to favor the most ethical players.

This article was written without AI.

[Thaura.AI](https://thaura.ai/home) was used for text's grammar review.**

References:

Amodei, D. (2026). Statement from Dario Amodei on our discussions with the Department of Wa. rAnthropic [online]: https://www.anthropic.com/news/statement-department-of-war

Alvarez, J. S. V. (2024). The risks and inefficacies of AI systems in military targeting support. Humanitarian Law & Policy. [online]: https://blogs.icrc.org/law-and-policy/2024/09/04/the-risks-and-inefficacies-of-ai-systems-in-military-targeting-support/

Baggiarini, B. (2026). The Pentagon strongarmed AI firms before Iran strikes — in dark news for the future of ‘ethical AI’. The conversation. [online]: https://theconversation.com/the-pentagon-strongarmed-ai-firms-before-iran-strikes-in-dark-news-for-the-future-of-ethical-ai-277198

Cape.co (2026). Claude AI Privacy Policy Explained: What the Fine Print Really Says. Cape Blog. [online]: https://www.cape.co/blog/claude-ai-privacy-policy

Cape.co (2026). Inside the Google Gemini Privacy Policy: Is Your Data Safe? Cape Blog. [online]: https://www.cape.co/blog/google-gemini-privacy-policy

Cape.co (2026). Perplexity AI Data Privacy Policy Explained [And How To Stay in Control]. Cape Blog. [online]: https://www.cape.co/blog/perplexity-ai-data-privacy-policy

Csernatoni, R. (2024). Governing Military AI Amid a Geopolitical Minefield. Carnegie Europe. [online]: https://carnegieendowment.org/research/2024/07/governing-military-ai-amid-a-geopolitical-minefield

Demarais, A. (2026). AI on the battlefield: How is the US integrating AI into its military? EuroNews — Tech News [online]: https://www.euronews.com/next/2026/03/06/ai-on-the-battlefield-how-is-the-us-integrating-ai-into-its-military

Del Rey, M. (2025). Amazon sued over Ring facial recognition, plaintiffs seeking $5M. USA Today. [online]: https://eu.usatoday.com/story/news/nation/2026/06/04/amazon-ring-lawsuit-facial-data-privacy/90404040007/

Duz, S. (2025). Gaza as a testing ground: Israel’s AI warfare. Seta Org. [online]: https://www.setav.org/en/gaza-as-a-testing-ground-israels-ai-warfare

Field, H. (2026). Anthropic refuses Pentagon’s new terms, standing firm on lethal autonomous weapons and mass surveillance. The Verge. [online]: https://www.theverge.com/news/885773/anthropic-department-of-defense-dod-pentagon-refusal-terms-hegseth-dario-amodei

Frias, L. (2025). See the Anduril drones that are taking AI-driven warfare to new heights. Business Insider. [online]: https://www.businessinsider.com/anduril-drones

Gonzalez, J. S. (2026). AI fever sparks an IPO race that threatens to change the balance of financial markets. El país — Economy & business. [online]: https://english.elpais.com/economy-and-business/2026-06-10/ai-fever-sparks-an-ipo-race-that-threatens-to-change-the-balance-of-financial-markets.html

Gould, L. (2025). Does AI really reduce casualties in war? “That’s highly questionable,” says Lauren Gould. Utrecht University News. [online]: https://www.uu.nl/en/news/does-ai-really-reduce-casualties-in-war-thats-highly-questionable-says-lauren-gould

Gvaryahu, A. (2026). These aren’t AI firms, they’re defense contractors. We can’t let them hide behind their models. The Guardian: Featured essay. [online]: https://www.theguardian.com/us-news/ng-interactive/2026/mar/15/ai-defense-warfare-companies

Human Rights Watch. (2021). Killer Robots: Survey Shows Opposition Remains Strong. [online]: https://www.hrw.org/news/2021/02/02/killer-robots-survey-shows-opposition-remains-strong

International Comitee of the Red Cross. (2025). UN Security Council: We cannot let AI be deployed on the battlefield without oversight and regulation. [online]: https://www.icrc.org/en/statement/we-cannot-let-AI-be-deployed-on-battlefield-without-oversight-and-regulation

International Energy Agency (2025). Electricity consumption by data centres, 2020–2035. Chart. [online]: https://www.iea.org/data-and-statistics/charts/electricity-consumption-by-data-centres-2020-2035

International Energy Agency (2025). Key Questions on Energy and AI. [online]: https://iea.blob.core.windows.net/assets/3179f7f8-01f6-4dd6-bffa-c9f7b73f1dc9/KeyQuestionsonEnergyandAI.pdf

Jeroudi, L. (2025). The Backlash Against Military AI: Public Sentiment, Ethical Tensions, and the Future of Autonomous Warfare. Trends Group — Research and Advisory [online]: https://trendsgroup.org/insight/the-backlash-against-military-ai-public-sentiment-ethical-tensions-and-the-future-of-autonomous-warfare/#_ftn3

Kali, H. (2024). The new ‘land grab’ for AI companies, from Meta to OpenAI, is military contracts. Fortune. [online]: https://fortune.com/2024/11/27/ai-companies-meta-llama-openai-google-us-defense-military-contracts/

Loizos, C. (2026). Mira Murati steps back into the spotlight, carefully. Techcrunch [online]: https://techcrunch.com/2026/06/04/mira-murati-steps-back-into-the-spotlight-carefully/

National Centre for Atmospheric Science(2025). Conscious use of AI: Practical tips and thoughts from environmental researchers. [online]: https://ncas.ac.uk/conscious-use-of-ai-practical-tips-and-thoughts-from-environmental-researchers/

Open AI. (2026). Privacy Policy. [online]: https://openai.com/en-GB/policies/row-privacy-policy/

Parreño, J. (2026). Open source models are good enough. Stop overpaying for intelligence you don’t need. Data Science Collective — Medium [online]: https://medium.com/data-science-collective/open-source-models-are-good-enough-stop-overpaying-for-intelligence-you-dont-need-3b533c716f55

Pistilli, G. & Luccioni Sasha. (2025). Ethics + Sustainability = Responsible AI. Hugging face, Community Blog and articles [online]: https://huggingface.co/blog/sasha/ethics-sustainability

Russell, B. (2018). Amazon pitched its facial recognition system to ICE. The Verge. [online]: https://www.theverge.com/2018/10/23/18013376/amazon-ice-facial-recognition-aws-rekognition

Singh, K. (2025). OpenAI wins $200 million US defense contract. Reuters. [online]: https://www.reuters.com/world/us/openai-wins-200-million-us-defense-contract-2025-06-16/

Spirling, A. (2023). Why open-source generative AI models are an ethical way forward for science. Nature. [online]: https://www.nature.com/articles/d41586-023-01295-4

Stahl, L. & Chasan, A. (2024). Kenyan workers with AI jobs thought they had tickets to the future until the grim reality set in. CBS News — 60 Minutes overtime. [online]: https://www.cbsnews.com/news/ai-work-kenya-exploitation-60-minutes/

Thaura AI. The Thaura Constitution. [online]: https://thaura.ai/constitution

Tillipman, J. (2026). Military AI Policy by Contract: The Limits of Procurement as Governance. Lawfare Media. [online]: https://www.lawfaremedia.org/article/military-ai-policy-by-contract--the-limits-of-procurement-as-governance

Toh, A. & Ayoub, E. (2026). Brennan Center for Justice — Research and Report. The Military’s Use of AI, Explained. [online]: https://www.brennancenter.org/our-work/research-reports/militarys-use-ai-explained

Transparency Coalition (2026). TCAI Guide: How to stop your images and data from being used to train AI. [online]: https://www.transparencycoalition.ai/news/tcai-opt-out-guide-how-to-stop-ai-from-using-your-images-and-data