Daily AI users who trust their employer least were about four times as likely to hide their methods

The four-times figure compares the two ends of the trust scale. Split 604 respondents into quarters and each holds about 150 people ^[1]. Eric Anicich, an associate professor of management and organization at USC Marshall who ran the survey with PhD student Jeslyn Brouwers ^[1], set the lowest-trust quarter against the highest ^[4]. Comparing extreme quarters gives a bigger ratio than a comparison across the whole scale would. The phys.org account does not give the withholding rate inside either quarter, a statistical test of trust against fear of job loss, or whether the work has been peer-reviewed. In absolute terms, roughly 200 of the 604 withheld something ^[2].

On job loss, the account is qualitative. It reports that some concerns centered on job loss and internal competition, but much of the hesitation came from mistrust of the organization ^[5]. "What separated the sharers from the hoarders was trust in the organization," Anicich said ^[3]. He described that as psychological safety, "whether you feel comfortable sharing what you know without worrying you'll be ridiculed or punished for it" ^[15].

Trust and withholding both come from the respondents' own answers. An employee who has already kept a technique back may rate the employer lower afterwards. A supervisor who punishes disclosure could push both measures down at once. Survey answers like these cannot separate those paths from the one Anicich describes, in which low trust produces the hoarding. The respondents were also heavy users by design, people on AI at work daily or several times a day ^[1].

Anicich said the reputational fear has support in other work. He cited research finding that "learning that a person used AI can lead observers to attribute less competence, motivation, effort, creativity, authenticity, or trustworthiness to that person" ^[8]. "Absent a clear signal that their organization values AI experimentation, withholding is the rational move," he said ^[7]. On replacement, he described what an employee gives up by sharing. "Right now, knowing how to actually get results out of these tools is a differentiator. It's what makes someone hard to swap out," Anicich said ^[9].

The employer loses sight of the method as well. "If your best analyst is fast because of a method nobody has seen, you can't train anyone else on it, and you don't really know what you're paying for," Anicich said ^[13].