Signals · Accountability
14% of staff reported an AI error and were told to stay quiet. Keep the record anyway
GoTo's survey of 2,500 workers finds AI errors going unchecked and unreported, so a dated record of what you caught is now worth keeping.
by Jo·4 min read·
New here? Start with the free AI Survival Kit →
14% of staff reported an AI error and were told to stay quiet. Keep the record anyway
0:00 / 6:28
This voice is generated by AI.
IT Brief reported on 1 October that GoTo, working with Workplace Intelligence, surveyed 1,250 full-time knowledge workers and 1,250 IT decision-makers across the US, Canada, the UK, Ireland, Germany, Austria, Switzerland, India, Mexico and Brazil. The headline figure is that 83% of employees fear they could be blamed or fired over an AI-driven mistake. The more useful figure sits further down. 14% of employees said they had reported an AI error to a manager or leader and were told to stay quiet about it. Meanwhile the national data still shows no break for the group most often named as exposed: an NBER working paper by Robert Fairlie and Jane Wu, issued in September, found that unemployment among recent US college graduates did not spike in summer 2026 compared with previous summers. The unemployment numbers for that group have not moved. The workflow has, and the accountability inside it is going missing.
The errors are already reaching customers
GoTo's IT leaders are past the hypothetical stage. A quarter said AI had already made mistakes that hurt customers or clients. 23% said it had caused financial harm to their company. 91% said they were concerned AI would make a mistake that affects their organisation.
The employee side shows how loosely AI output gets checked. 55% said they do not always review or double-check AI outputs, even for high-stakes tasks. 43% said they had used AI outputs they believed were low quality or suspected contained errors or fabricated information.
The survey links some of that behaviour to pressure. 60% of employees feel pressure to use AI to increase productivity. 47% say AI use is considered in their performance reviews. 42% would rather risk making errors with AI than look "slow" or "replaceable". That is a workforce under pressure to use AI, with nearly half seeing it counted in reviews.
It is tempting to treat this as a policy gap the employer will close. The figures say do not wait for it. Only 24% of small businesses and 36% of midsize companies have an AI policy. Even among enterprises the figure is only 53%. Everywhere else, nobody has written down what gets checked or who answers for it. 84% of workers said their employer could do more to encourage responsible AI use.
Silence is the actual failure mode
Fear of blame reads like an anxiety problem. It is better read as a price signal. The survey shows accountability being offloaded in three directions at once. 26% of employees said using AI lets them avoid personal accountability at work. 25% said it feels safer to blame AI than to admit an error, and 17% said they had already blamed AI for a mistake of their own. And 31% believe there is unspoken pressure at work to trust AI and stay quiet about its mistakes.
Put those together and the picture is an organisation running AI where nobody owns the output. The tool gets blamed, the report gets buried, and the next error ships the same way.
That gap is where the value now sits. A system can draft the contract clause, the forecast or the customer reply. It cannot sign for it. The person who will say "I checked this, here is what was wrong, here is what I changed" is doing the part of the job that cannot be delegated to the tool. AI does the routine work. You do the thinking, and owning the result is part of the thinking.
The time to do it exists. Employees who use AI told GoTo they save an average of 2.3 hours a day. Spend part of that on the check and the record, and you hold the layer your organisation has not yet staffed.
Your Next Move
-
Start an AI error log this week. One line per incident: date, tool, task, what was wrong, how you caught it, what it would have cost if it had shipped. Five minutes an entry. Within a month you have evidence instead of an impression, and evidence is what changes a manager's mind.
-
Write your own review line before your employer writes one. List the outputs you always check in full: anything going to a customer, anything carrying a figure, anything with a legal or financial consequence. List the ones a spot-check covers. Send it to your manager as your working standard. In a firm without a policy, the first written standard tends to become the standard.
-
Raise errors as fixes, with the record attached. Bring the log entry and the check that would have caught the problem. If you are told to stay quiet, keep the dated entry anyway. If AI use counts in your review, a log of catches is the proof of judgement that a usage count never shows.
I build systems like the one publishing this site. → Work with me
About the author
Jo
Jo runs The War Room: one signal a day on how AI is changing work, and what to do about it.
Sources
Get the Briefing
Want more intelligence like this?
The weekly briefing, free — and the AI Survival Kit with it.
More in Signals
See allSignals · Capability
Siemens CEO Roland Busch says AI could transform society in 7 to 8 years, and AI skills will decide who gains
3 min read
Signals · Power
Windows 11 is recording which AI coding agents you run, and in one US test the upload setting was on
3 min read
Signals · Labour Market
Pay moves to the bottleneck. A careers list that is really a map of who owns the work
3 min read