As a consumer, I’ve been conditioned to prize the real deal, the genuine article. To me, that feels at odds with generative AI, a technology infamous for making things up. But AI blunders like dressing people of color in Nazi uniforms are just one of many reasons to be distrustful.

Look, even the folks who build AI don’t know how it works. It promotes harmful stereotypes, and its very creators fret about AI-induced human extinction. 

No wonder almost eight out of 10 Americans don’t trust businesses to use AI responsibly. For bosses as well as workers, an AI trust gap persists too. 

For brand trust, AI introduces a whole new risk category, says Beena Ammanath, leader of trustworthy AI and technology trust ethics with Deloitte. But the biggest risk is not using AI, Ammanath warns. “If you don’t use it, you’re going to lose your competitive advantage.” 

So, where to start building trust? Philipp Herzig, chief AI officer of software giant SAP, tells customers using AI to begin by thinking about their core beliefs and what’s important to them. “The technology moves super fast,” says Herzig, whose company released a comprehensive AI ethics policy back in 2018. “But values shouldn’t move fast.” 

Living those values means giving people the final say. “AI is a very, very useful tool,” Herzig says. “But at the end of the day, as it relates to decision-making, it needs to be done by a human.”

For oversight, SAP has an AI ethics governance body. At one point, the company considered adding an AI feature to its SuccessFactors HR solution that would let users generate questions for job interviews, partly based on a candidate’s CV. But its AI ethics group decided that even after removing personally identifiable information (PII), the data privacy risk of using CVs was too high.

Deloitte identifies six dimensions of trustworthy generative AI. It should be fair and impartial, transparent and explainable, safe and secure, accountable, responsible, and respectful of privacy. Tall order.

Making AI transparent and explainable calls for strong communication, both inside and outside the company. First, ensure everyone on the team knows enough about AI to ask the right questions and flag potential problems, Ammanath says. “Building that base-level AI literacy for your employees is something that I’m seeing happening across enterprises today.”

Second, be straight with the public. For instance, a company should clearly label AI-generated content, or disclose in plain English how it uses the data gathered by its AI customer support system. “It’s no longer about just putting out legal [or] really complex documents but explaining it in a language that your customer can understand,” Ammanath explains.

A business must also define “who is accountable when AI produces inaccurate results, or when it goes racist or misogynistic,” says the author of Trustworthy AI: A Business Guide for Navigating Trust and Ethics in AI.

As AI advances and evolves, companies need to watch out for new “side effects” that could impact trust, Ammanath notes. “We are truly entering that phase of being agile, being nimble on adapting to these new technologies, along with the risks that come with it.”

Trust me, AI will keep us all guessing.

Nick Rockel
nick.rockel@consultant.fortune.com

IN OTHER NEWS

Here’s some advice
Warren Buffett doesn’t hold back in his latest annual letter to shareholders. Besides warning that markets are turning into a casino, the Berkshire Hathaway CEO channels his late right-hand man, Charlie Munger, by declaring that financial pundits should “always be ignored.” Why? Being one is “like finding gold and then handing a map to the neighbors showing its location,” Buffett says.

Fast talk
Should we take Elon Musk at his word? The Tesla boss claims his company’s next-gen Roadster, which just wrapped engineering, can lap Formula 1 race cars by going from zero to 60 in less than a second. In the works since 2017, the new EV could ship next year.

Generation Gig
Gen Z has little faith in the traditional nine-to-five career, a new international survey reveals. Seven out of 10 respondents said they’re freelancing now or plan to go that route soon, with one-third reckoning it means less chance of being laid off or replaced by AI. For Gen Z, self-employment is also about having more control over their lives.

Banking crisis
When it comes to RTO, Deutsche Bank workers don’t trust that their employer is doing the right thing. Many have spoken out against the bank’s post-pandemic demand that come June, all staff must be in the office at least three days a week. Up next: more talks with labor reps.

TRUST EXERCISE

“Fraud and graft exist everywhere in the world, albeit to different degrees, but the need for Ukraine to cleanse a culture of everyday ethical malfeasance is critical to its immediate future. It will take time.”

Along with a war on its hands, Ukraine has a trust problem, argues James P. Moore Jr., founder and CEO of the Washington Institute for Business, Government and Society. For Western business executives and other observers, the perception that the country is corrupt could weaken support for the fight against Russia.

Moore calls on Ukraine to step up its anti-corruption efforts, which have made strides under President Volodymyr Zelensky but failed to stamp out bribery and other bad behavior. To that end, his organization is helping launch a nationwide ethics education program spanning business, government, and academia.