Cybersecurity expert Morgan Wright says old-fashioned human caution—not sophisticated detection technology—offers the best defense against AI-powered financial fraud. Generative AI tools keep getting better at mimicking voices and writing convincing messages, and scammers can exploit personal vulnerabilities faster than software can catch them. Regulators and tech companies are struggling to keep pace.
Wright has trained federal and state law enforcement, including teams at the FBI and National Security Agency. He cited a stark example: Deborah Del Mastro, a California mother, got a call from what sounded exactly like her daughter, who claimed she was being kidnapped. “It’s my daughter’s voice,” Del Mastro recalled. “And she’s having a panic attack.” Convinced her daughter faced immediate danger, Del Mastro transferred more than $5,000. She later discovered the call was a scam using voice-cloning technology.
“When you can fool a mother, that’s how good these things are getting,” Wright said. The scammer needed only a few seconds of audio, often pulled from social media, to create a realistic imitation. The technology keeps advancing. Detection tools struggle to adapt.
The Technology Outpaces Detection
Wright said software-based defenses face an inherent disadvantage. “We hear about all the different models from xAI, Anthropic, OpenAI and Gemini. They change models constantly,” he explained. “So, when the model changes, how effective is [detection software]? We don’t know because this stuff is moving so fast. What you see today will be drastically different tomorrow.”
Adversarial states like China and Russia deploy AI globally, while cybercriminals usually target ordinary Americans with simpler but highly effective techniques. They focus on psychological manipulation, not technical sophistication. The scams work because they exploit human emotion—urgency, fear, family bonds—rather than relying on technical deception alone.
Wright’s advice harks back to Edgar Allan Poe: “Believe nothing that you hear and only one half that you see.” He recommended pausing before acting after receiving a suspicious call or message. “We’re not storming the beaches at Normandy; you’re not going to get arrested,” he noted. “No law enforcement agency sends you an unsolicited text saying you’re about to get arrested.”
Public Concern Outpaces Administration Action
A Reuters/Ipsos poll conducted over six days ending Monday found significant public anxiety about AI risks and the government’s response. The survey of 4,506 U.S. adults, including 3,526 registered voters, found that 57% of registered voters believe the Trump administration hasn’t taken AI risks seriously enough or at all. Roughly a third of self-identified Republicans share that view. Likewise, 54% said Congress hasn’t adequately addressed the threat.
The concerns run broad and deep. Eighty-four percent of voters see AI as a threat to American workers—a higher share than those citing illegal immigration (60%), foreign worker competition (74%), or large corporations (73%) as threats to workers. Sixty-two percent believe AI could get out of control and threaten humanity’s future, roughly in line with a Reuters/Ipsos poll from August of last year.
A bipartisan majority backs stricter government regulation of AI: 80% of Democrats and 61% of Republicans support tighter rules. Fifty-six percent favor limits on data-center construction, but Republican support falls to 43%, compared with 69% among Democrats.
Administration Priorities Diverge From Public Sentiment
The Trump administration has chosen a different approach, prioritizing AI development and data-center construction as essential to U.S. competitiveness against China. Trump has repeatedly argued that the U.S. technology industry could fall behind Chinese counterparts without aggressive support for AI growth and infrastructure.
A White House official responding to the poll results cited the administration’s new AI task force, saying it’s responsible for ensuring “America continues to lead the world” in the technology. “The Trump administration’s priority is to ensure that there is a booming economy so that workers have many opportunities and experience limited volatility from any short-term disruptions,” the official said in response to employment concerns.
Daniel Kokotajlo, a former OpenAI staffer who left the company in 2024 because he felt it wasn’t behaving responsibly, offered a different perspective. “That’s an important concern but it’s not top of mind for most Americans, nor should it be,” he said about competitiveness arguments. He argued that the more immediate worry centers on job losses and threats to human health and safety.
Public concern has sharpened over military applications. Fifty-seven percent of voters said the U.S. government shouldn’t use AI to choose targets for military strikes, up from 49% in a Reuters/Ipsos poll conducted in August 2025.
Why This Matters:
The gap between public anxiety and government strategy reveals a fundamental tension in AI policy. Most Americans—including a substantial share of Republicans—want tighter regulation and more caution, while the administration is betting that economic growth and technological leadership justify faster deployment. Wright’s emphasis on personal vigilance over technological fixes suggests citizens may need to become their own front line of defense, regardless of regulatory approach. The scam epidemic also shows how quickly AI capabilities can outpace detection systems, leaving software solutions alone inadequate. If detection stays behind the curve, individuals and institutions must build resilience through behavioral discipline rather than expecting technology to solve the problem. This dynamic could reshape how businesses, government agencies, and families approach digital security, shifting responsibility downward even as public opinion demands government action.