AI can offer considerable advantages, from simple machine learning that seems to know what you meant to type, to more complex algorithms that can predict health care needs and detect patterns in climate change. It is now routinely used across the tech spectrum and often kicks in without the user even realizing that it is there.

However, it also poses significant threats to privacy, data management, and the prospect of machine ‘learning’ that leads to unwanted surveillance, racial profiling or discrimination. And it is hard to know the true state of affairs due to a lack of disclosure about companies’ AI activities.

This lack of information was one of the reasons why the Active Ownership team was only able to successfully conclude four out of five cases in the engagement program that ran from 2019 to 2022. The other five cases were transferred to the team’s SDG Engagement theme to further engage these companies on their societal impact.

“Through our engagement, we learned that companies are gradually aligning internal practices to principles of responsible AI,” says engagement specialist Danielle Essink. “Many companies formalized AI principles that address topics like inclusion, fairness and transparency.”

“Additionally, companies are increasingly pursuing a collaborative approach by actively contributing to initiatives that aim to advance responsible governance and best practices. These types of initiatives play a decisive role in guaranteeing trustworthy AI across the industry.”

“However, ethical principles on their own do not ensure the responsible development and deployment of AI. Businesses require robust governance mechanisms to effectively implement their principles.”

A major stumbling block is the lack of disclosure about what companies are actually doing to address concerns, along with their willingness to engage in the first place. Much of the AI technology and how it is implemented on different platforms is still shrouded in secrecy.

“In our engagement, we observed that transparency around governance and implementation remained low, as most companies’ public disclosures lacked clarity about how such principles translate into practice, and which checks and balances are in place,” says engagement specialist Claire Ahlborn, who also worked on the program.

“After talking to the companies, we learned about the specifics of the implementation, which then gave us the confidence to close some of the objectives successfully. The engagement results of this theme are, therefore, highly correlated with the company's willingness to set up constructive dialogues.”

The International Data Corporation’s Worldwide Artificial Intelligence Software Forecast 2022 said the worldwide AI market is set to show compound annual growth of 18.6% from 2022 to 2026 alone.

Yet the potential benefits come with risks that are not yet fully explored, let alone understood, Essink says. To achieve the full potential of AI, companies need to manage the associated risks that come with the development and use of the technology, including human rights-related risks.

“Given the speed at which AI is being developed, there is no doubt that in the next few decades, this technology will transform our economy and society in ways we cannot imagine,” Essink says.

“This type of growth represents massive opportunities for AI to contribute to positive changes, such as detecting patterns in environmental data, or improving the analysis of health information.”

“At the same time, AI could cause new problems or aggravate existing ones if companies do not have enough understanding of the risks associated with these technologies. For example, using AI algorithms for profiling can have discriminatory effects, such as credit rating algorithms disfavoring people from certain ethnic backgrounds, or those living in certain areas.”

“Similarly, AI can be used for surveillance – in public spaces but also in the workplace – putting the right to privacy at risk. This shows a growing need for the responsible governance of AI systems to ensure that such systems conform to ethical values, norms, and the growing number of AI regulations.”

Such regulatory moves and policy proposals have already been launched by governments, ethics committees, non-profit organizations, academics and the EU. In April 2021, the European Commission issued the AI Act. It sets out clear requirements and obligations regarding the specific uses of AI for developers, deployers and users.

The proposal identifies four regulatory categories based on the level of risk. At the high end, AI systems identified as high-risk, such as CV-scanning tools that rank job applicants, will be subject to strict obligations including enhanced risk management processes and human oversight. AI systems with limited risks will remain largely unregulated.

“This growing legislative pressure around AI could pose serious regulatory risks for companies that are not well prepared to conform with the rising obligations,” says Ahlborn.

“Meanwhile, the alignment of AI technologies with ethical values and principles will be critical to promote and protect human rights in society. As a result, we will continue our engagement work with a selection of companies in the tech sector under our ‘Sustainable Development Goals (SDG) engagement’ theme.”

“These dialogues have a strong focus on human rights and societal impact, and highlight topics like misinformation, content moderation and stakeholder collaboration. We will focus on how companies can contribute to SDG 10 (Reduced inequalities) and SDG 16 (Peace, justice and strong institutions) by safeguarding human rights in the development and use of AI and promoting social, economic and political inclusion.”