Nobody at work wants to admit they use AI

Among workers already anxious about layoffs, 60.7% have used AI to take over tasks that used to belong to a coworker, and 62.8% never told their manager AI was doing the work, according to ResumeBuilder's April 2026 survey of 1,000 full-time US employees. At companies that had already conducted layoffs, that concealment climbed to 74.3%.

Employers are just as quiet. The AFL-CIO's April 2026 AI and Work Survey, conducted by David Binder Research, found that 94% of workers say they should know when AI is being used to monitor their work. Just 7% say their employer has actually told them.

Workers are hiding how much AI is doing their job. Employers are hiding how they're monitoring it. The manager caught in between is left running performance reviews on a picture neither side will complete.

The real reason AI use stays hidden

Grant Wycliff, president of Carolina Georgia Sound, subscribed to a personal Claude account two years ago. "It started as something I was just trying out and grew into a daily tool before I ever thought to make it official," he says. "The difference in quality of work with and without it was obvious, and others around me noticed the difference." A multi-site technology assessment that normally took him a full day landed in a client's inbox within hours.

"I wasn't really certain of how a client would react to finding out that the deliverables they were paying experts to get were done with the help of AI," he says. "I told them that I had made my research and writing procedures more efficient with the help of some improved tools." He never named the tool. "This was true, but it was also intentionally vague," he says. "Admitting that there was some sort of tool behind this was certainly challenging and a plain yes or no didn't cover it."

37% of employees say AI tools have made them appear more competent than they actually are, and 47% have stayed quiet rather than admit they don't know how to do something. Among those who stayed quiet, 50% say they were expected to figure it out on their own, 49% didn't want to appear incompetent, and 29% didn't feel safe admitting it. Employees already falling behind on learning are more than twice as likely to say AI made them look more competent than they are (66% vs. 29%) and more than twice as likely to have stayed quiet (76% vs. 31%), according to TalentLMS's Learning Debt Report. AI is making the underlying skills gap easier to live with, and harder to see.

Workers hide AI use to protect their perceived competence, says Paula Pinzon, an ISO/IEC 42001-certified AI governance auditor and professor in AI for working professionals. "The dominant driver I see is not fear of punishment but protection of perceived competence, disclosing AI feels like admitting the output isn't 'theirs,'" she says.

Nobody defined the line

Gabriella Goddard, founder of Brainsparker, coaches leaders on how their teams can use AI. One colleague preferred a personal Claude or ChatGPT subscription over the company's own tool because it "gave them access to the latest models and consistently produced better results than the organization's approved AI platform." Goddard doesn't read it as an attempt to bypass policy. "They genuinely wanted to do better work," she says. "The challenge was that the gap between their personal AI experience and their workplace AI experience had become so wide that using their personal account felt like the best practical option."

At Suff Digital, COO Kriszta Grenyo says her team had never defined acceptable AI use. “No one was intentionally crossing a line because we didn't establish a line to cross in the first place,” she says. “Since we had never defined what was appropriate and what wasn't when using AI, our employees were left to determine that on their own.” The issue surfaced when the team discovered some employees were using AI to polish emails while others were pasting sensitive information into prompts.

Grenyo's team isn't unusual. Only 36% of employees say their workplace has clear AI policies and approved tools, and 1 in 10 describe their organization's AI environment as "the Wild West," according to a 2025 survey by Laserfiche.

"In audits [of companies' AI governance practices], monitoring opacity is usually governance immaturity rather than strategy," says Pinzon. "Companies deploy monitoring tools faster than they write the transparency policies an AI management system requires, so there is literally nothing official to disclose, and legal teams prefer silence over documenting an unpoliced practice."
62% of US adults had heard nothing at all about employers using AI to monitor and evaluate workers, and just 6% had heard a lot about it, according to Pew Research Center.

At Pivot Creative Media, a team member started testing Claude on the company's network without approval, and founder Tyler Desjardins caught it in the logs. “When we spoke to this person, they had no idea we could see any activity on our network, which made me realize how we had neglected to share information about monitoring with our employees,” Desjardins says. At the time, “there was never any formal communication” about the monitoring, he says, and Pivot did not yet have an employee handbook. After the incident, employees learned through internal meetings and Slack that the company was “monitoring network activity and why.”

Managers can see the output, not the capability

Managers' own AI proficiency scores are barely higher than those of the employees they manage, according to Section's AI Proficiency Report. Managers who aren't proficient with AI themselves can't effectively guide their teams on how to use it: 65% either set no expectations at all or encourage AI without holding anyone accountable, and only 7% require AI use and tie it to performance evaluations.

At Microsoft, Minh Pham, a Senior Finance Manager who leads AI Implementation Initiatives for Windows Consumer Finance, has integrated Copilot and AI-assisted workflows into his team's daily work. “Microsoft gives us the tools and expects us to use them, but specific guidance on how to weigh AI-assisted output in a performance review doesn't really exist yet,” Pham says. “Managers are mostly building their own judgment calls team by team and for each different project.”

Once Copilot workflows went live on his team, executive narratives that used to take a day were ready in an hour. "The problem wasn't catching someone hiding AI use, as we encourage increasing usage," Pham says. "It was figuring out how much of that jump reflected their own judgment versus the tool." Early in the rollout, that ambiguity had a real cost. “There were plenty of errors when my direct reports didn't fully understand the context behind the numbers and blindly trusted AI. I started asking people to walk me through their reasoning, not just hand me the output, to see what they actually understood," he says.

Pham encourages his team to automate work when employees can still defend what they produce to clients and stakeholders. “What's harder is the person who leans on it as a crutch and stops building independent judgment,” he says. “Flagging that without discouraging the tool's use entirely is a real balancing act.”

He judges promotions by that standard too. “Volume and speed have reduced its importance in judging the success of an IC.” What matters more, he says, “how they can come up with new analysis and storytelling skills. That's harder to observe than output, so promotion conversations now lean more on how someone thinks and what new workstreams they can create by synthesizing inputs from different stakeholders and databases.”

Most managers aren't doing what Pham does. 41% of employees say their role has evolved faster than their company's ability to train them, and 62% use workarounds when they lack the skills or training to complete a task, according to TalentLMS's Learning Debt Report. 29% say they have delivered work they could not fully explain if asked how they did it, while 50% say AI helps them complete tasks even when they do not understand the process behind them.

Employees falling behind on learning are nearly six times more likely to make errors that proper training could have prevented, 47% versus 8%, TalentLMS found. More than a quarter, 28%, say their manager does not know how often they struggle with the skills their job requires.

Dr. Cassidy Blair, a licensed clinical psychologist and founder of Blair Wellness Group, worked with a team lead who never told their manager they were using AI to pre-process data before passing it to the next stage. “Their team's output went up, and their manager started praising the entire group,” Blair says. When she asked why the team lead had kept the AI use quiet, they told her: “Because the final output belongs to them, not the tool. They still put in the work to come up with those outputs.”

“When a majority of AI-assisted work is submitted as fully human, performance data stops measuring human capability,” Pinzon says. “You end up promoting prompt-engineering skill while recording it as domain mastery, and misdiagnosing training needs across entire teams.”

Make AI visible, then judge the thinking

68% of employers now have a formal policy governing AI use, up from 38% a year earlier, according to Littler's 14th Annual Employer Survey. Pinzon says disclosure rises when employees know AI use is expected and employers are equally clear about how they monitor it. "Frameworks like ISO/IEC 42001 force exactly that reciprocity," she says. Those policies ask workers to show how they used AI and tell them what workplace monitoring is in place.

The employee Desjardins caught testing Claude still uses AI, but with a caveat. "[He's] definitely become more open with us after we indicated we weren't interested in monitoring productivity but rather security issues," Desjardins says. "[He]'s been asking which tools we deem appropriate for him to use. He understands that we are more concerned with security over what he is doing [with AI]."

90% of HR leaders say AI has fundamentally redefined what "high performance" means, but only 42% have updated goal-setting or review criteria to reflect it, according to the Betterworks 2026 State of Performance Enablement Report, a survey of 2,387 HR leaders, managers, and employees.

Edward Tian, founder of AI-detection company GPTZero, has changed how he evaluates his own team's work. "I no longer use the phrase 'prove AI usage' in my evaluations of employees. I instead ask them to take me through the steps of how they solved this," he says. "Strong employees will be able to articulate the steps taken to create a work product, irrespective of whether they utilized AI tools. Weak employees will struggle to articulate any steps taken to arrive at their completed work."

At Tally Workspace, co-founder Laura Beales evaluates whether employees can turn AI-assisted analysis into useful business judgment. "I'm evaluating if [my employees are] making good decisions, growing the business, and taking things off my plate," she says. "AI can't do that part for you."

“Workers aren't hiding AI because they fear the tool; they're hiding it because nobody told them what disclosure costs. Employers aren't hiding monitoring out of malice; they're hiding it because they never wrote the policy,” Pinzon says.