Few applications of artificial intelligence have captured the public imagination, and skepticism, like autonomous vehicles. Here, machine learning models must navigate not just unpredictable road conditions and erratic human drivers, but also intense regulatory and media scrutiny. Even after years of remarkable technical progress, public trust is not universal. For many, the idea of an AI-powered car handling rush-hour traffic still feels closer to science fiction than everyday reality.
What does it take to build AI systems the public will trust? Few are better positioned to answer than Chinmay Jain, Senior Technical Product Manager at Waymo and IEEE Senior Member. Over a fifteen-year career spanning Waymo and Google, Jain has built products at the intersection of cutting-edge innovation and user trust. At Waymo, he helped steer the Driving Behavior team from early ride-hailing pilots to a fleet logging more than 250k trips per week. His experience underscores a hard truth: technical breakthroughs and marketing campaigns mean little unless AI behavior consistently meets real-world expectations.
Going Off Script: Real-World Testing
In controlled environments, AI can perform near perfectly. In the real world, however, every mile driven introduces new variables, ranging from weather shifts to unexpected driver behaviors, that models must handle without explicit pre-programming. A striking example came in 2018, when researchers showed that placing stickers on a stop sign causes object detection systems to misclassify them. Although industry practices have since evolved to address vulnerabilities like these, the episode serves as an early example of the complex world AI encounters outside the lab.
“Scaling AI into the real world requires much more than model accuracy metrics,” Jain notes. “There’s always the question of how the system handles ambiguity, especially when human safety is involved. There’s zero room for error.”
Under Jain’s leadership, Waymo prioritized an evaluation-first product culture. Instead of relying solely on aggregate accuracy scores, his teams created evaluation sets designed to surface rare but consequential edge cases, such as navigating emergency scenes or managing complex construction zones. By prioritizing consistent behavior across these edge cases, they sharply reduced disruptive “unwanted stops,” an ambiguous but highly visible problem for riders helping clear the path for Waymo’s expansion into major cities, where today the service completes more than 250,000 paid passenger trips per week.
Meanwhile, shifting federal policy—including recent Department of Transportation revisions aimed at strengthening U.S. competitiveness—has placed even greater pressure on companies to prove safety and public reliability. Jain notes that the global race for autonomy depends on both technological progress and sustained public acceptance.
Meeting Users Halfway: Designing AI for Public Trust
Outperforming human drivers statistically isn’t enough to win public trust. A single awkward encounter—a sudden brake at a green light, a hesitation at a busy crosswalk—can outweigh thousands of uneventful miles in a rider’s mind. According to Pew Research surveys, concerns over AI have risen steadily since 2021, often magnified by viral stories of rare but unsettling incidents.
“Trust is emotional, so theory and testing will only take you so far,” Jain says. “To earn that last mile of acceptance, AI needs to behave in ways people intuitively understand and expect, over and over again.”
Recognizing this, Jain and his teams treat public officials as key collaborators. Working closely with entities like the California DMV, they incorporated regulatory feedback into product design, shaping vehicle behavior to meet not just legal standards but evolving social expectations. Internally, Jain emphasized building cross-functional teams where engineering, safety, legal, and policy experts worked side by side. This approach helped anticipate public concerns early and built credibility during expansion into complex cities like San Francisco.
The takeaway is simple: building AI for public spaces requires a deep understanding of perception, societal context, and ethical responsibility. It mirrors why explainable AI has become a priority in sectors like finance and healthcare, where trust hinges as much on how decisions are made and communicated as on the outcomes themselves.
Toward a Culture of Trustworthy AI
As autonomous vehicles move from pilot projects to everyday services, Jain believes the challenges they’ve addressed will become instructive across all AI-driven industries: Trust will come from a culture of transparency and direct engagement with public concerns. It has to be earned regularly and openly.
“We’re asking people to trust something they can’t fully see or control, or always understand,” Jain says. “Building that trust is always a work in progress.”
Real-world exposure also helps shift perceptions. According to J.D. Power’s U.S. Mobility Confidence Index, consumers who have personally ridden in autonomous vehicles report nearly double the trust levels compared to those who have not. Firsthand experience tempers emotional reactions to isolated incidents and counters sensational headlines.
Across mobility, finance, healthcare, and personal technology, the pattern holds: AI will earn trust not through promises or statistics, but through consistent, predictable, understandable behavior—decision after decision.
