Corporate leaders across the United States are racing to adopt artificial‑intelligence (AI) tools, but a new Zapier‑commissioned survey shows many employees are still unsure how to protect company information when using these systems. The study, conducted by research firm Centiment between May 13 and May 21, 2026, polled 1,005 U.S. workers at companies with at least 500 employees that provide paid AI access for business tasks.
Key findings on data handling
Overall, 89% of respondents admitted to entering some form of company data into an AI tool. More striking, 79% said they have entered highly sensitive information such as human‑resources records, login credentials, API keys, or personally identifiable information (PII) belonging to customers or employees.
While many workers recognize the risk, 85% say they usually consider whether information is safe before sharing it with AI. Still, 36% are not confident they know what is permissible, and 72% worry their organization could face data‑exposure or compliance problems because of AI use.
Why the risk matters
Hackers are already exploiting unsecured AI agents to harvest credentials and other valuable data. When sensitive information is entered into ungoverned tools, it becomes a prime target for theft, potentially compromising not only the company but also its customers and partners.
Workarounds and policy gaps
About one‑quarter of respondents (26%) reported that their employer blocks or restricts certain AI applications, yet they still accessed those tools using personal devices or accounts. A similar share (24%) admitted to bypassing restrictions by copying data into personal tools or switching browsers. Among those using personal AI accounts for work, 43% prefer to keep that usage hidden from their employer.
Training makes a difference
The survey highlights a clear link between formal training and confidence. Eighty‑two percent of workers who feel fully confident about AI data handling have received dedicated training, compared with only 46% of those who feel uncertain.
Recommendations for leaders
To protect valuable corporate data, the report urges executives to establish clear governance policies that define approved AI tools, permissible data types, and consequences for non‑compliance. Effective measures include:
- Designating approved AI platforms and disabling unauthorized ones.
- Implementing access controls, logging, and monitoring to track data flow.
- Providing regular, mandatory training on safe AI practices.
- Embedding security guardrails directly into AI workflows so the safe path is the easiest path.
While eliminating every risk is impossible, a structured approach can dramatically reduce exposure and give employees the confidence to use AI responsibly.
About the survey
The study surveyed U.S. workers at companies with 500 or more employees that have paid‑plan access to AI services such as ChatGPT Enterprise, Claude for Business, or major cloud AI APIs. Data are unweighted, with a margin of error of roughly ±3% at a 97% confidence level.
Zapier produced the story, and Stacker reviewed and distributed it.
Original reporting: KRDO (Colorado Springs metro) — read the source article.
