Today's Editorial

Reshape the governance structures of AI companies

Relevance: GS Paper III

Why in News?

The modern corporate governance regimes in capitalistic and neo-capitalistic economies have traditionally favoured the theory of shareholder primacy. 

More about the issue:

• In modern corporations, the objectives of profit generation and wealth creation for the shareholders and investors take primacy over other objectives of the business including the objective of public good. 
• In contrast, there have been proponents of a stakeholder benefit approach of corporate governance, which seeks to maximize the benefits of all stakeholders.
• In recent years, corporations with ostensibly alternative governance models, leaning towards stakeholder capitalism have become more common. 
• Corporations are increasingly getting involved in products, technologies and services that cannot be driven solely on the objectives of profit making and have a greater social objective. 
• Generative Artificial Intelligence (AI) is one such instance, where corporations are seeking alternative governance structures to balance the objectives of generating profit with that of greater social responsibility.

Risks posed by AI advancements:

• Data access issues:

• • The development of AI technologies requires access to data, which may, in turn, accelerate the ability to utilize personal information to undermine privacy. 
  • For instance, Meta was asked to pause its plans to train its large language models using public content shared on Facebook and Instagram in the European region over concerns raised by the Irish privacy regulator. 
  • In addition to this, it has been noted that human prejudices may find their way into AI systems and lead to algorithmic biases with harmful results.
  • Recently, Amazon discontinued using a recruiting algorithm after it discovered that it was plagued with gender bias. 
  • Moreover, researchers at Princeton University conducted an experiment where they used AI software to analyze and link words and found that European names were perceived as more pleasing than their African-American counterparts. 
  • These examples demonstrate how AI can perpetuate existing biases and create inequality with respect to opportunities, and access.

• Purpose versus profits:

• While these companies started out with alternative models, when there was a clash between the company’s goals of purpose and its profit-generating machinery, the monetary interests won. 
• OpenAI, the creator of ChatGPT, found itself embroiled in a corporate governance debacle last year when the non-profit board of the company fired the CEO of the company due to concerns about the rapid commercialisation of AI products at the cost of compromising user safety.
• This debacle has raised questions on the viability of public benefit corporate structures in the technological industry, which rely on capital infusion from shareholders and investors with deep pockets, to fund research and innovations.
• From these recent events, it is evident that even in this new age of public benefits corporation, the purported public benefit may be nothing more than disguised profit seeking.
• The present accountability structure is not sufficiently strong to protect against this amoral drift, where the social objectives of a corporation are often subsumed by the broader profit-driven goals as the market enables unrestricted corporate control. 

Steps Taken by Some Big companies:

• Recently, Amazon discontinued using a recruiting algorithm after it discovered that it was plagued with gender bias.
• To counter the risks posed by AI advancements, OpenAI, and Anthropic, have resorted to structures with public good and developing responsible AI as core objectives leading to creation of public benefit corporations. 
• For instance, Anthropic is governed by a structure called Long-Term Benefit Trust. This trust is composed of five financially disinterested members who have the authority to select and remove a portion of Anthropic’s board. 
• Similarly, OpenAI was incorporated as a non-profit, but it transitioned into a hybrid design by incorporating a capped profit-subsidiary to support its capital intensive innovation.

Workable strategy:

• Policymakers need to employ innovative methods of regulating corporations involved in developing AI-based products which balance these conflicting interests.

• From a strictly economic perspective, this can be done by targeting three key areas: 
• Enhancing long-term profit gains of corporations from adopting a public benefit purpose.
• Incentivising managerial compliance of such purposes.
• Reducing compliance costs of adopting such purposes.
• This would require framing ethical standards for the governance of AI product companies, along with providing adequate regulatory backing through reforms in corporate governance norms. 
• It is important for the creators of AI to act responsibly towards all stakeholders.

Conclusion:

With the increasing involvement of AI in multiple spheres of life, it is imminent that governance models promoting the ethical development of AI for generating profits need to be adopted.