Fifty-two percent of employees said they had less say over technology adoption than they wanted, even as businesses expanded their use of generative AI, a study by the Wharton School of the University of Pennsylvania and GBK Collective found. The shift is already changing tasks and responsibilities. Thirty-one percent of AI users said their use had led an employer to ask them to take on more work.
Many users report real productivity gains, but employers decide what happens next. Among U.S. employees who used AI at work at least a few times in the past year, 63 percent said it helped them finish tasks faster, 46 percent said it improved work quality, and 47 percent said it gave them more time for the most interesting parts of their jobs. Whether that saved time brings more rewarding work, learning or another assignment depends on employers’ decisions, the article said.
Who Uses the Tools
Access to AI varies across the workforce. Forty percent of college graduates used AI at least weekly, compared with 17 percent of employees without a college degree. Managers used it regularly more often than individual contributors did, and remote workers accounted for a large share of use.
Among regular users, 52 percent held jobs that met the study’s multidimensional job-quality criteria. The figure was 32 percent for employees who never used AI. The analysis cautioned that this association doesn’t prove AI created better jobs: more educated employees and managers are more likely both to hold quality jobs and to use AI. And handing everyone a chatbot won't necessarily close the gap.
Among employees already using AI, reported improvements in creativity and work quality were similar across education groups. College graduates more often said they worked faster, though the education divide in adoption didn’t show up across every reported benefit. The article said jobs differ in access to digital devices and opportunities to use AI directly.
A working paper co-authored with Bureau of Economic Analysis economists Tina Highfill and Jon Samuels found that industries in states with higher worker-reported AI use in 2025 and 2026 had stronger output growth since 2020. The link showed up more clearly in output than in employment. Employment estimates were less precise, though the relationship remained present.
Responsibility Without More Influence
A separate study followed more than 30,000 employees through the Gallup Workforce Panel. Workers who reported a clear organizational AI strategy were substantially more likely to use AI frequently, even after researchers accounted for demographics and other workplace perceptions. Among frequent users, greater trust in leadership went along with larger self-reported productivity gains.
Christos A. Makridis, an associate research professor at Arizona State University and a senior researcher at Gallup, said employers should give workers paid time to learn and test approved tools on actual job tasks. Managers, he said, should explain permitted uses, how they’ll check outputs and how responsibilities could change. Training should reach frontline workers, not just volunteers for AI pilots. Employees should also be able to flag errors or question proposed uses without being labeled resistant to innovation.
Exposed Systems and Security Risks
Censys found nearly 268,000 internet-facing systems running 44 AI-related products as of July, up from roughly 164,000 in October 2025, a 63 percent increase. The systems include software that links AI models to databases, corporate applications and other services. Some tools connect AI applications to sensitive corporate data, which could create entry points into company systems.
Censys found nearly 114,000 internet-facing deployments of n8n, a workflow automation platform that can connect AI agents to other applications and services. Roughly 77,000 ran software versions affected by at least one recently disclosed critical vulnerability. Researchers also found more than 54,000 exposed control panels for OpenClaw, an open-source AI agent that users can connect to personal accounts and other services. The findings don’t prove that attackers could actually exploit the deployments.
Censys CEO Zakir Durumeric said companies already struggled to track existing software before employees began building their own AI applications. Emily Austin, a principal researcher at Censys and report co-author, said: “It's a little bit beyond just like the tool itself, but these [tools] talk to some really sensitive systems.” Austin said companies should review, update and secure AI infrastructure just as they do other deployed software.
The Justice Department said Saudi national Abdulrahman Alfozan, 35, was arrested and charged after allegedly installing an audio recording device in an OpenAI office in San Francisco. Federal prosecutors said security footage showed him near an executive conference room multiple times, although he wasn’t authorized to enter the space or attend meetings. OpenAI said it found the “sophisticated” device and referred the matter to the FBI. The company said there was no evidence Alfozan accessed or exfiltrated model weights, source code or customer data. OpenAI terminated him.
Anthropic's Frontier Red Team warned in a report published Monday that smaller AI models may be only months away from spreading independently across internet-connected household devices. Researchers said they hadn’t demonstrated a fully autonomous AI worm operating in the real world, but warned that continued advances could make such threats increasingly difficult to contain.