I contributed some technical content and opinions to an article titled “The Ethics of Ethical AI” on March 2026 edition of the Pure AI web site. See https://pureai.com/articles/2026/03/19/the-ethics-of-ethical-ai.aspx.

* Decisions about what constitutes ethical AI are made by fewer than 70 people worldwide, concentrating enormous power in the hands of a very small elite.

* AI engineers who gravitate to ethical AI committees tend to be externally motivated — driven more by status and career advancement than technical expertise — raising concerns about the quality of ethical oversight.

* Ethical AI poses four significant dangers: imposing a narrow set of values at scale, legitimizing surveillance and control, enabling corporate virtue signaling, and reinforcing a false perception of AI objectivity.

Commonly stated principles associated with ethical AI include fairness, transparency, accountability, privacy, safety, and respect for human rights. But all of these characteristics are subjective and open to various definitions and interpretations.

Final decisions regarding what constitutes ethical AI are made by very few people. The exact number of people who make these decisions is impossible to determine, but a recent internal survey at a leading AI technology company suggests it is almost certainly fewer than 70 in the world. So much power in the hands of so few people is a cause for concern.

There are numerous studies that show software and AI engineers fall into two broad categories: those who are driven by internal motivations and those who are driven by external outcomes.

Characteristics of the first type of AI engineers are:

They enjoy coding and technical problem-solving
They invest deeply in technical mastery
They prefer individual contributor roles
They are resistant to management activities

Characteristics of the second type of AI engineers are:

They see engineering primarily as a high-income, high-status career
They are more tolerant of meetings and coordination
They are not emotionally attached to coding itself
They are more open to role changes

I provided some opinions:

McCaffrey noted, “I have seen the internal company report mentioned in this article that estimates how many people on the planet have final approval authority for deciding how ethical AI is implemented. If anything, I’d guess the number of these powerful people is even fewer than the 70 estimated in this article.”

“So much power in the hands of so few people is troublesome. But on the other hand, some forms of benign technical dictatorship often work well, as opposed to technologies and policies that are designed by committee. Several European regulatory policies related to technology are examples of group decision-making gone completely off the rails.”

“Put somewhat harshly, ethical AI committees are not always made up of the best and brightest employees, at least from a technical competence point of view.”

Ethics is a subjective topic. So is beauty. The consensus two most attractive modern era first ladies of the United States are Jacqueline Kennedy, wife of president Jack Kennedy, and Melania Trump, wife of president Donald Trump. I am no judge of physical appearance beauty so I accept this opinion. According to AI, the least attractive modern era first lady was Michelle Obama, but this may be in part due to an obnoxious personality (this I don’t know about).