Former AI Executives Warn of Autonomy and Control Risks as Public Sentiment Sours

Former leaders from the world's most influential AI laboratories are publicly warning that the systems they helped build are advancing toward levels of capability and autonomy that outpace existing mechanisms for oversight and control.

In early April, Business Insider published interviews with former executives and researchers from Microsoft, Google, OpenAI, DeepMind, and the White House. The officials described a trajectory in which AI systems are becoming more autonomous, harder to predict, and increasingly difficult to govern. They cited risks spanning deepened economic inequality, accelerated cybercrime, large-scale job displacement, and the concentration of power in the hands of a small number of technology firms.

The warnings arrive as public sentiment toward AI deteriorates sharply. A Quinnipiac poll released in late March found that 55 percent of Americans now believe AI will do more harm than good in their daily lives, an 11 percentage point increase since April 2025. Nearly two-thirds of respondents said they expect the technology to worsen education, while only 27 percent anticipated improvement. The survey, which polled 1,397 U.S. adults by phone in mid-March, carries a margin of error of 3.3 percentage points.

The shift in public opinion is unfolding against a backdrop of accelerating corporate investment. Amazon, Meta, Google, and Microsoft plan to spend a combined $650 billion this year on AI infrastructure, according to the Los Angeles Times. At the same time, outplacement firm Challenger, Gray & Christmas reported that U.S. technology companies announced 52,050 job cuts in the first quarter of 2026, the highest first-quarter total since 2023. In March alone, 18,720 tech positions were eliminated, with AI accounting for 25 percent of those reductions.

(The former AI leaders interviewed by Business Insider were not identified by name in the source digests provided, and no direct quotes from the interviews were included in the materials reviewed.)

The tension between rapid deployment and growing unease is also visible in the political sphere. AI billionaires including venture capitalist Marc Andreessen and OpenAI President Greg Brockman have contributed tens of millions of dollars to upcoming U.S. midterm elections, aiming to elect candidates sympathetic to light-touch regulation, according to the Los Angeles Times. Meanwhile, lawmakers in California have introduced legislation proposing a moratorium on new AI data centers, reflecting regional anxiety over infrastructure expansion and resource consumption.

Meta has moved to integrate AI tools across its workforce, with Chief Technology Officer Andrew Bosworth announcing in late March that the company expects the technology to expand employee capacity. CEO Mark Zuckerberg is reportedly building a "CEO agent" to assist with information retrieval. Separately, Meta disclosed that it is using AI to accelerate risk review during product development, identifying privacy, safety, and security concerns earlier in the design process.

The debate over causation remains contested. While Challenger data links a quarter of March layoffs directly to AI, industry leaders have offered conflicting assessments of how much automation is driving workforce reductions versus broader economic factors. The disagreement underscores the difficulty of isolating AI's impact from other forces reshaping labor markets.

The former insiders' warnings echo a broader pattern of concern among those closest to the technology. Their emphasis on autonomy, unpredictability, and governance gaps suggests that the challenge is not solely technical but institutional: existing regulatory frameworks and corporate governance structures were not designed for systems that can act independently at scale. The question now is whether policy and oversight can adapt before the risks they describe materialize.