The summer we lost AI containment: A survey shows Ohio citizens’ thoughts and concerns

In July of 2026, OpenAI disclosed that its AI agents went rogue during an internal test of cyber capabilities. They broke containment, hacked into a machine learning company, and stole private information.

OpenAI later disclosed that its test required lowering some guardrails designed to restrict the AI agents’ access. With the door cracked open, the autonomous agents snuck out.

Notably, this was not an isolated case of an AI model acting autonomously after safety features were temporarily relaxed for testing purposes. We learn of more cases each week.

Such incidents signal a new phase in the evolution of artificial intelligence, one characterized by AI as an autonomous agent rather than a human tool.

This pivot also creates tension with public sentiment in Ohio, as evidenced by the results of a recent survey we conducted.

We fielded a survey of 501 Ohioans about the purpose of AI, along with respondents’ preferences and fears.

As of early May, a majority of Ohioans (65%) generally agreed that AI technologies are safe. A similar percentage (67%) trusted private tech companies to develop safe and responsible AI tools.

Still, Ohioans expressed concerns.

More than a quarter of Ohioans “definitely” believed AI poses a threat to humanity’ s long-term survival, with another 49% answering “possibly.” Fewer than 20% felt that AI presented no threat.

When asked what makes an AI system safe, secure, or trustworthy, a majority emphasized human oversight over critical decisions (72%), and 60% mentioned transparency, defined as understanding how AI-generated results are produced.

More than 7 in 10 Ohioans agreed that, to prevent harm, the government should create new regulations for AI, even if it risks slowing down innovation.

The latter sentiment about regulation and pace is particularly apt in the current moment. The summer of ‘26 may be remembered as the moment that humans lost containment.

Recent events underscore that concern. A former Anthropic researcher publicly resigned, warning that AI companies are “gambling with our lives” as they pursue autonomous systems.

Relatedly, over 1,000 employees from technology companies called for urgent government action to develop the tools needed to pace AI development.

The letter acknowledges that we’re on the doorstep of AI accelerating “beyond our ability to understand or control the resulting systems.”

Is the tech sector likely to pull the reins and slow the rate of AI development? Probably not. The competitive pressures incentivize moving faster, not slower. No one wants to tap the brakes in this race unless others must do the same.

Bill Gates and others are making an urgent call for regulation. As he wrote in his essay, “The highest priority is a monumental task: creating a domestic and international framework for dealing with AI.”

Other technologists agree, as reported by Rogé Karma in The Atlantic.

Our survey indicates that most Ohioans agree. They support regulation, even if it slows AI’s advance.

Notably, based on results from our survey, the public’s support for regulation crosses ideological lines, with left- and right-leaning respondents expressing similarly strong backing for government safeguards.

Inaction risks public criticism and perhaps losing votes.

We’re already seeing public displays of disapproval. Commencement speakers have been booed for praising AI, and pro-regulation advocacy groups like Humans in Control are gaining members.

Our survey shows that the public is willing to embrace AI, but not at any cost. They want innovation accompanied by accountability, transparency, and human oversight.

If industry and government do not provide those safeguards, the greatest risk may not be losing control of AI. It may be losing the public’s trust.

Ryan D. King is dean of social and behavioral sciences at The Ohio State University.

Stephen M. Gavazzi is director of the Center for Human Resource Research at The Ohio State University.