FDA Seeks Generative AI Device Framework as Pew Documents Rising Public Job Anxiety

Pew Finds Rising Concern and Deepening Youth Pessimism on Employment

A Pew Research Center survey conducted from June 22 to June 28 among 3,488 U.S. adults revealed that 52 percent of respondents feel more concerned than excited about artificial intelligence in daily life. That proportion stood at 37 percent in 2021. By comparison, 9 percent express more excitement than concern, while 37 percent report equal levels of both.

The shift appears most pronounced among adults under 30. Within that demographic, 55 percent express greater concern than excitement, an increase from 31 percent in 2021. Furthermore, 73 percent of respondents under 30 expect artificial intelligence to decrease the total volume of jobs in the United States over the next two decades, up from 61 percent recorded in 2024.

Pew emphasized that the findings reflect public sentiment rather than an economic forecast. Only 5 percent of all participants anticipated that artificial intelligence will create more jobs. While the responses do not establish that massive job losses are certain, they demonstrate that skepticism now dominates employee expectations.

FDA Solicits Input on Generative AI Medical Device Competency

The Food and Drug Administration opened a public consultation on August 18 covering risk assessment, premarket evaluation, postmarket monitoring, foundation models, and autonomous agentic systems in medical hardware and software. The agency set an October 19 deadline for comments from manufacturers, clinicians, and researchers.

The FDA emphasized that its discussion paper remains a non-binding inquiry rather than formal regulatory guidance. It introduces potential oversight ideas without altering established approval thresholds or determining whether the regulator requires additional statutory authority from Congress.

Among the proposed concepts is a two-axis risk model alongside a competency evaluation procedure modeled after physician licensing. Under this approach, developers would complete non-clinical benchmarking followed by clinical validation proving that a tool operates as intended. The agency noted that a medical model may produce articulate responses while suffering from poor calibration, low reproducibility, demographic bias, or performance drift following underlying model modifications.

New York Examines Automation Pressures on Administrative Workers

On August 19, Axios reported that New York Governor Kathy Hochul convened initial listening sessions through the state FutureWorks Commission. The sessions focus on women whose daily employment has already experienced disruption from artificial intelligence tools.

According to state data cited by the governor's office, women occupy 84 percent of New York's administrative, clerical, and customer-service positions. State officials have identified these occupational segments as facing high exposure to automated task execution.

The FutureWorks Commission comprises 20 representatives from private industry, organized labor, higher education, and state policy agencies. The panel holds a mandate to submit policy recommendations by the end of the year to protect workers' economic stability while supporting productivity gains. Gathering direct testimony from administrative personnel aims to uncover operational friction and undocumented workflows that formal corporate planning documents often overlook.

Operational Boundaries Replace Broad Performance Metrics

The convergence of regulatory inquiry and workforce wariness illustrates that system deployment requires verifiable guardrails. A successful enterprise implementation requires clear boundaries: explicit documentation of tasks an automated model may handle, defined circumstances requiring human intervention, and transparent metrics for ongoing oversight.

When employees believe automation serves solely to eliminate positions, willingness to document institutional knowledge or report edge-case errors declines. Organizational leaders must treat workforce trust as a functional operational metric. This requires publishing explicit criteria on how automated systems reach conclusions, detailing which duties are changing, and establishing clear mechanisms for workers to dispute automated outcomes.

Meaningful worker involvement demands structural support rather than retrospective public forums. Standard practices include offering paid time for staff to evaluate tools, whistleblower avenues to report errors without fear of discipline, worker ownership over personal performance data, and advance notification before software changes influence compensation or job design.

Procurement Standards Shift Scrutiny to Vendor Continuity

Institutional purchasers of generative systems now encounter technical evaluation challenges similar to those raised by the FDA. Software buyers must determine whether updates to base foundation models invalidate prior validation studies and whether internal teams can independently audit log files.

Procurement teams must verify how vendor platforms identify subtle performance drift across distinct user populations and whether external subjects possess rights to challenge system determinations. Without explicit contractual assurances and clear operational records, the deploying enterprise absorbs all legal and reputational exposure when an automated tool malfunctions in production.