How to write AI ethics policies in the workplace

If there is one general thread across recent research on AI in the workplace, then there is no conclusive view of how people use technology and how they feel about the orders to do so.

Language learning models can be used to draft policies, generated AI can be used to create images, and machine learning can be used to use for predictive analytics. Senior Capterra analyst specializing in trends in the HR industryI told HR Diving by email.

Still, there is a clear lack of what tools to use, and for a wide range of applications in the market, Bahr said. Although organizations implement these tools, “use policies often get confused by employees,” Bahr told HR Dive in an email.

The consequences can be unethical or even illegal. The use of AI can create data privacy concerns, violate state and local laws, and raise claims. Identity-based discrimination.

Compliance and culture hold hands

AI ethics policies primarily address compliance, but culture is equally important. If employers could explain the reasons behind AI rules, “employees feel empowered by AI rather than being blackmailed,” Bahr said.

“By ensuring human surveillance and telling that AI is a tool to support workers rather than turnovers, we create an environment in which employees use AI responsibly, not just employees,” Bahr added.

Kevin Frechette, CEO of AI software company Fairmarkit, highlighted a similar theme in his advice to HR professionals building AI ethics policies.

The best policy answers two questions: “How does AI help teams do their best work?

“If you can't answer how AI improves someone's day, you probably aren't ready to write a policy,” Frechette said in an email.

Many policy conversations prioritize technology on behalf of workers themselves, he said.

Consider industry-specific issues

There are additional layers to consider in the industries involved in creating AI tools. Bahr pointed to a study from Capterra last year that revealed software vulnerabilities were the biggest cause of data breaches.

“AI-generated code or vibe coding can present a security risk, especially when AI models are trained in public code and accidentally replicated existing vulnerabilities into new code,” Bahr explained.

AI disclosure policies should address security risks, provide training to develop internal review guidelines for AI-generated code, and promote secure coding practices, Bahr said.

For businesses involved in creating content, AI disclosure may be required, Bahr said, and it is necessary to address how workers are responsible for the final product or outcome.

“This policy not only shows the public that human input is involved in publicly available content, but also establishes employees with a responsibility to comply with the disclosures they require,” Bahr said.

“Beyond fact-checking, this policy must address the use of intellectual property in public AI tools,” she said. “For example, entertainment companies need to be clear about using the voices of actors to create new lines of dialogue without permission.”

Similarly, software salespeople can explain to their clients how AI is used in their company's products. For example, the use of customer data could also be part of our disclosure policy.

The policy is in place. What now?

As AI technology continues to evolve, experts say employers need to remain flexible.

“Static AI policies become obsolete before the ink dries,” Frechette of Fairmarkit said. “We treat it like a living playbook that evolves with technology, regulation and workforce needs,” he told HR Dive in an email.

According to Frechette, HR also needs to continue testing and update its AI policies regularly. “It's not about perfecting it on the first day,” he said. “It's about making sure it's still relevant and effective after six months.”