How Leading Organizations Are Implementing a Global AI Ethics Framework

Executives have already integrated artificial intelligence (AI) into many business processes. However, as they continue exploring the most viable applications, these leaders must uphold ethical practices while using the technology. Establishing a global AI ethics framework that includes fairness, data privacy, accountability and transparency helps them do that.

Fairness

Many AI tools are only as good as their training data and processes. Since humans influence those things, bias is a common problem in many AI models. Unfortunately, leaders often use them to make decisions that could drastically impact many people’s lives or job prospects. Business leaders should create ethical frameworks that ensure fairness and include processes for minimizing bias.

Amazon uses AI in its hiring processes. However, since the results could be life-changing for applicants, the company prioritizes fairness in these applications. For example, a group of computer scientists continually monitors AI-suggested candidate matches to check for fairness. Recruiters review that content to verify that the affected applicants genuinely seem like good fits for long-term company success.

Fairness is also an essential component in education-based AI applications. Leaders and professionals in this field should not use tools trained on copyrighted material without the creators’ permission and check that AI results do not perpetuate biases. Prioritizing fair outcomes is also important in cases where educational institutions establish policies for student AI usage.

For example, many colleges forbid learners from using AI to write class assignments, and professors use scanners to check for AI content. However, these scanners are less than 80% accurate in detecting AI usage on average. Establishing processes that allow parties to appeal cheating allegations maintains fairness since AI tools are not infallible.

Data Privacy

Data privacy is a multifaceted part of a comprehensive AI ethics framework because decision-makers must consider what they are trying to protect and create appropriate policies. Some companies have restricted what employees can do with generative AI, such as stipulating that they cannot create prompts containing sensitive or proprietary information.

However, the data privacy component may also need to reflect location-specific laws, ensuring that AI usage complies with regulations. For example, LinkedIn does not train its AI tools on data from users living in places with strong privacy laws, including the European Union, China and Hong Kong. However, people living elsewhere deserve privacy protections, too. The best thing for them to do until such regulations exist is to become familiar with what information the AI tools consume and why. 

Then, people may be able to take certain actions to protect themselves. For example, Meta representatives have said it does not train its AI models on private messages or posts. After learning that, users could decide to tweak settings so they do not have any public-facing content for the company to use.

Accountability

Although AI can improve many business processes, it is an imperfect technology, and those using it should remain accountable when unexpected consequences arise. Globally known publisher The Guardian demonstrated how to do that when it stipulated how its journalists would and would not use AI. Part of that effort involved three principles guiding its use of the technology. These principles benefited readers, the organization and journalists.

The outlet’s leaders promised to tell readers if pieces contained substantial AI-generated elements and would only include such material after applying human oversight. They would also consider the impact on journalists by avoiding models that contained material used without permission. Although organizational policies will vary, these are some good examples that executives can use in their businesses.

Transparency

Transparency about businesses’ uses of AI can secure customers’ trust. How and when companies should disclose this information to stakeholders is a much-discussed topic, and experts have proposed valuable ideas. In one case, a panel of AI professionals supported mandatory disclosure, with 84% agreeing or strongly agreeing with that approach. However, the individuals varied in their ideas about implementing such a policy.

Additionally, some experts suggested emphasizing the possible risks of using AI, similar to how pharmaceutical companies must list notable side effects in television commercials. However, others believe these disclosures could become unnecessarily burdensome, especially for smaller companies.

Decision-makers should gather information from numerous affected parties before proceeding with a single approach. That feedback can give them balanced perspectives to shape the outcomes.

Responsible AI Usage

These examples show how ethical AI practices can build trust, mitigate risks and enable innovation. Detailed and thorough ethical frameworks can shape current and future business decisions, leading to responsible AI usage.