Ethical AI Was Delayed by Choice, Not Necessity

The claim that “ethical AI cannot exist” is wrong. Ethical AI has been possible all along; what has been missing is the willingness of the institutions building, buying, and governing AI to accept slower growth, lower margins, and real constraints.

The arc of the story

In 2017, AI ethics was still mostly a warning signal. By 2019, the OECD had adopted its AI Principles, establishing the first intergovernmental standard on AI. UNESCO followed in 2021 with its global ethics recommendation. The U.S. published the Blueprint for an AI Bill of Rights in 2022, NIST released its AI Risk Management Framework in 2023, and the EU AI Act finally entered into force in 2024. The key point is not that nothing was said; it is that formal ethics were avoided by intentional acceleration.

This record shows a repeated pattern: warnings came first, principles came later, and binding limits came last—if they came at all. That is not proof that ethical AI is impossible. It is proof that powerful actors preferred speed and profit over caution and accountability.

In the United States, the Blueprint for an AI Bill of Rights in 2022 and the executive order on safe, secure, and trustworthy AI in 2023 both showed that ethical principles could be translated into governance goals: notice, human alternatives, protection from algorithmic discrimination, testing, oversight, and accountability. NIST’s AI Risk Management Framework gave industry and government a concrete structure for doing the work. But these efforts were late, uneven, and dependent on voluntary uptake or executive action rather than durable, universal law.

Globally, the OECD AI Principles in 2019 and UNESCO’s Recommendation on the Ethics of Artificial Intelligence in 2021 created a shared ethical vocabulary. The problem was not the absence of norms; it was the absence of binding force. Norms do not stop harmful deployment when the institutions that profit from deployment are free to ignore them.

The clearest counterexample to the “ethical AI cannot exist” claim is the EU AI Act. Whatever one thinks of its details, it proves the point: ethical limits can be written into law. Prohibited practices, high-risk obligations, and compliance duties show that a society can decide to constrain AI before harm becomes the price of admission. That is what ethical AI looks like when governments choose to make it real.

So who is accountable? First, business: companies had the earliest power to slow deployment, test systems, disclose risks, and refuse unsafe scale. They usually did not. Second, governance: regulators and lawmakers often preferred voluntary principles to enforceable rules. Third, global institutions: they articulated standards but lacked teeth. Citizens matter too, but they are not the primary authors of the failure. The people with the most control are the people most responsible.

The objection that ethics “cannot exist” is really an objection to accountability. It is easier to call ethical AI impossible than to admit that ethical AI would require giving up some immediate advantage. But that is exactly why the phrase matters: if a system cannot survive even modest restraint, then the problem is not ethics. The problem is the system.

Where ethics could have been instituted

The biggest missed decision points were these:

• 2017–2019: before frontier-model scale-up, governments could have imposed baseline testing, disclosure, and audit requirements.
• 2019: when the OECD Principles were adopted, states could have turned them into enforceable law rather than soft coordination.
• 2021–2022: when UNESCO and the U.S. Blueprint appeared, Congress and regulators could have converted ethical language into mandatory rights, redress, and procurement rules.
• 2023: when NIST published the AI RMF and the U.S. executive order was issued, agencies could have required independent audits, incident reporting, and pre-deployment review across sectors.
• 2024 onward: with the EU AI Act, the main question shifted from whether ethics could be defined to whether it would actually be enforced.

Who is accountable

Business: companies moved first and fastest, and in doing so set the tempo for everyone else. They chose deployment before restraint.

Governance: governments often knew the risks but preferred voluntary frameworks, advisory principles, and pilot programs over binding obligations.

Global governance: bodies like OECD and UNESCO clarified norms, but norms without enforcement do not stop harm.

Citizens and civil society: public pressure mattered, but ordinary users were never positioned to compensate for weak law or weak corporate restraint.

So the accountable parties are not just “AI” or “society.” The primary blame lies with firms and governments that had the authority to slow deployment and did not.

Conclusion

Ethical AI is not a contradiction nor is it an impossibility. It is a governance choice. The fact that it has so often been postponed does not prove it cannot exist; it proves that powerful actors preferred immediate advantage to responsible delay.

References

1. OECD. AI principles. https://www.oecd.org/en/topics/ai-principles.html
2. OECD. The state of implementation of the OECD AI Principles four years on. https://doi.org/10.1787/835641c9-en
3. NIST. Artificial Intelligence Risk Management Framework (AI RMF 1.0). https://doi.org/10.6028/NIST.AI.100-1
4. NIST. AI Risk Management Framework. https://www.nist.gov/itl/ai-risk-management-framework
5. The White House. Removing Barriers to American Leadership in Artificial Intelligence. https://www.whitehouse.gov/presidential-actions/2025/01/removing-barriers-to-american-leadership-in-artificial-intelligence/
6. European Commission. AI Act. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
7. European Commission. AI Act enters into force. https://commission.europa.eu/news-and-media/news/ai-act-enters-force-2024-08-01\_en
8. European Commission. European AI Office. https://digital-strategy.ec.europa.eu/en/policies/ai-office
9. UNESCO. Recommendation on the Ethics of Artificial Intelligence. https://www.unesco.org/en/artificial-intelligence/recommendation-ethics
10. U.S. White House Office of Science and Technology Policy. Blueprint for an AI Bill of Rights. https://www.whitehouse.gov/ostp/ai-bill-of-rights/