Singapore Survey Finds Senior Managers More Willing To Act On An Unverified AI-Cloned CEO Voice
The findings concern hypothetical responses, while Singapore cybersecurity guidance advises checking unusual requests through another channel.

Senior managers in a Singapore workplace survey were more likely than clerical workers to say they would comply with an urgent request in an AI-cloned chief executive's voice without checking it, highlighting the tension between authority and verification.
In research released by KnowBe4 on Oct. 13, 35% of senior managers said they would act on such an unverified request, compared with 16% of clerical workers. The senior-manager share was about 2.2 times the clerical share, calculated from the reported percentages.
Censuswide surveyed 500 Singapore office workers who use computers or email for work between Aug. 6 and 11, 2026. The findings were released during Singapore International Cyber Week and GovWare.
The survey also found that 36% of respondents were not confident they could distinguish a message from their chief executive from an AI imitation. Half said they would feel uncomfortable questioning a senior executive's instructions, while 52% feared negative consequences from doing so.
A Survey Of Intentions, Not An Attack Test
The results describe what respondents said they might do in a hypothetical situation. They do not measure how many employees actually approved a fraudulent payment or how many businesses suffered a breach. Treating the percentages as an observed incident rate would give the research a meaning its method does not support.
The comparison between job groups is also narrower than a finding about every manager in Singapore. The published release gives the overall sample size but does not specify the number of respondents in each seniority subgroup. It therefore does not provide the information needed to calculate the precision of that comparison independently.
KnowBe4 sells security awareness products, giving the publisher a commercial interest in the subject. That does not invalidate the responses, but it is relevant context when separating the reported survey findings from claims about which workplace interventions will be most effective.
Verification Through A Separate Channel
The Cyber Security Agency of Singapore's deepfake advisory, issued in March 2024, recommends assessing the source and context of suspicious content and verifying a request through another communication channel. Its guidance remains directly relevant to a voice that sounds familiar but carries an unusual instruction.
For an unexpected message apparently from a colleague, that can mean contacting the person using a known number or a separate established channel. The purpose is to check the request with the person it claims to represent, rather than relying on the sound or appearance of the incoming message.
The agency also describes visual and audio anomalies as possible warning signs. Those clues can prompt further scrutiny, but the absence of an obvious flaw is not proof that a message is authentic. Verification is a distinct step from making a judgement about how realistic the recording sounds.
Authority Can Complicate That Check
A request attributed to a chief executive carries organisational weight as well as a claim about identity. The survey's discomfort and fear responses point to a separate issue from detecting synthetic audio: whether an employee feels able to pause an instruction from someone higher in the hierarchy.
That makes the procedure for checking a request relevant to the entire organisation. If verification depends on an individual deciding whether to challenge a superior, the response can vary with confidence, seniority and the perceived urgency of the message. A known process can give staff an ordinary operational reason to make the check.
The research provides a snapshot of those stated attitudes in Singapore workplaces. The agency's guidance supplies a practical verification principle. Together, they direct attention to how a request is authenticated and authorised, as well as to whether its voice sounds convincing.

