Keeping AI Safe, Transparent, and Accountable

As AI increasingly becomes part of the fabric of business, government, and daily life, the organizations adopting it face a growing array of complex ethical questions.

How do you ensure the safe, transparent, and accountable adoption of this transformative technology? How do you prevent unforeseen negative consequences and harm to vulnerable groups?

AI governance turns these questions into tangible action.  

“Good intentions by themselves aren't good enough,” says Dave Prakash, director of AI governance at Booz Allen. “You need well-thought-out policies and processes to identify hidden risks.”

Maintaining a high standard of AI governance is both a leadership expectation and a guiding North Star at Booz Allen, and there’s a good reason why. “As the leading provider of AI services to the federal government, Booz Allen sets the standard for excellence in AI governance,” says Erin Sheppard, Booz Allen deputy general counsel.

“In the AI we develop, procure, deliver to our clients, and use internally, we have to make sure that trustworthiness is built in,” says Geoff Schaefer, Booz Allen chief AI ethics advisor who leads the firm’s responsible AI practice.

Dave, Erin, and Geoff discuss AI governance: why it’s so important, the investments and thought leadership behind Booz Allen’s approach, and why a broad range of perspectives—including those from their own diverse backgrounds—is vital.

Erin calls AI governance “both strategic and tactical.” Geoff describes it as “the gap between Aristotle and algorithms,” making sure philosophical principles make their way into specific decisions. When it’s lacking, says Dave, “organizations are either paralyzed with fear and trepidation, or they’re full steam ahead—build, build, build, thinking ‘I’m sure everything will work out.’ Until it doesn’t.”

Dave reminds hesitant organizations that nothing worth doing is without risk. For the full-steam-ahead crowd? “AI governance helps you move faster. It’s a seat belt, not a speed bump, on your journey,” he explains.

So, what exactly does AI governance look like in practice?

Within Booz Allen, “it’s a framework in which employees know what to do, where to go, and who to speak to as they interact with and build AI tools and solutions,” says Erin. “For example, our Responsible AI Guiding Principles set forth our clear expectations, aligned to the firm’s purpose and values. We also have an AI Policy and an AI Governance Playbook, both of which provide tactical, tangible guidance.” 

Within the larger AI governance ecosystem, Booz Allen is an active voice in the broader global dialogue. One example is the Responsible AI Top-20 Controls, developed by the Responsible AI Institute, which Booz Allen joined earlier this year. This collaboration gives Booz Allen and its clients access to leading assessments, benchmarks, and certifications aligned with global standards and emerging regulations.

The controls are aimed at making AI governance “practical, tangible, and useful,” says Geoff. “The idea is if an organization doesn't know where to start or what to do with AI adoption and governance, start here.”

Dave’s advice is to prioritize the people aspect—incentives and education—as they move forward. “If people aren’t motivated to understand the risks, they won’t participate in the governance process.”

People play an important part in creating and maintaining a governance framework as well.

“We recognize the inherent need for a collaborative approach,” says Erin. “AI governance is technical, it’s legal, it’s contractual. It’s a range of different things, so it’s important to bring the right stakeholders into the conversation.”

Geoff cites the importance of involving employees on the front lines—from developers to program managers—in the process, to “connect the dots” between what’s happening daily with the broader principles of AI governance.

Diverse backgrounds and perspectives are key to making it all work.

“You need to be thinking about a range of reasons when managing bias or other aspects. The worst thing you can do in this area is to be single-minded or have tunnel vision,” says Erin.

The Responsible AI team’s own backgrounds reflect this philosophy.

With degrees in law and public policy, Erin came to Booz Allen from private practice as a procurement lawyer. “I wanted to work for an organization that was mission-driven and at the forefront of solving our clients’ greatest problems,” she says.

Geoff’s career path evolved from finance jobs in government and the private sector to a role with the Carnegie Council for Ethics and International Affairs. “For me, it was the right time and place, and the more I started doing that work, the more I fell in love with it,” he says.

And Dave is uniquely accustomed “to bringing disparate worlds together and looking at problems from an outside perspective.”

“I started out as a doctor, then joined the Air Force as a pilot after 9/11,” he shares. After military service that involved work as a test pilot, flight surgeon, and evaluator of new weapons, tactics, and systems, business school followed. The opportunity to help launch an AI team at a large health insurance company taught Dave about building high-stakes AI systems in healthcare and the associated ethical risks.  

“In my current role, all the dots connect,” Dave says. “All my experiences are coming together to help me lead an unprecedented program of enormous consequence. The biggest takeaway so far: You don’t talk about values; you must demonstrate them.”

Doing so involves implementing the same steps and standards the team advises to internal clients throughout the firm—and living by these standards.

“Our AI governance is built not just to support the business we have today, but the business that we expect to become,” says Erin.