The ethics of generative AI tools from major industry players do not appear to differ significantly from one another, according to an analysis of current practices. The ethical concerns associated with generative AI primarily revolve around the development of these models, particularly how the data used for training is obtained, and the ongoing environmental impact of these technologies. Generative AI systems require massive amounts of data, and decisions made by developers to acquire this data are often considered questionable and lack transparency. Even purported “open source” models often do not disclose their training datasets.
There have been objections from creators, including authors, artists, filmmakers, and social media users, who do not wish their content to be used as training data for AI. AI companies often bypass the consent of these creators, justifying this by claiming that obtaining consent from all the creators would be too burdensome and would stifle innovation. Even when companies establish licensing agreements with major publishers, the data obtained through these agreements represents only a small fraction of the total data used.
Some developers are attempting to develop models that fairly compensate creators for the use of their work in AI training, but such projects remain niche and are not yet mainstream. Additionally, the environmental consequences are significant, with generative AI tools requiring much more energy to operate compared to non-generative counterparts. While there are potential methods to reduce energy consumption, such as the energy-efficient model proposed by DeepSeek, major AI companies are more focused on advancing development rapidly rather than improving environmental sustainability.
Addressing the concern about making AI more ethical rather than simply more powerful, some efforts are underway to embed ethical guidelines into AI development. For instance, Anthropic is employing a “constitutional” approach for its Claude chatbot, aiming to impart core values into the machine.
Confusion often arises from the language used to describe AI functionalities, such as “reasoning” and “chain-of-thought,” which can blur the distinction between human and machine capabilities. However, AI does not possess genuine reasoning abilities; these terms are merely descriptive of how algorithms process information. Ethical considerations for AI should ultimately focus on the intentions behind user interactions, potential biases in training data, and how developers program responses to controversial topics. The goal should be to develop more ethical practices in AI creation and user engagement, rather than solely enhancing AI’s intellectual capabilities.
