How to handle a candidate's inappropriate use of 'AI' during an interview?

I'm a little surprised that other answers did not emphasize what I think is the most important consideration here. It's why I would emphatically say you would be completely justified in rejecting this candidate: He lied by passing off the AI generated answer as his own. This was in the context of an interview trying to evaluate his skills and qualifications for the job. Yes, it may have been better if you'd given clear direction at the beginning of the interview that no outside resources should be used, but that is really not the issue.

He has shown himself to be untrustworthy. He's willing to cheat to get the job. There's no reason to think he won't cheat or lie on the job if he thinks it would be to his advantage.

Is it reasonable for us to reject this candidate for his use of AI during the interview?

Yes. If you had calibrated the difficulty of the interview so that no outside tools were needed, but the candidate had to use outside tools to solve it, then that tells you information about the capabilities and limitations of the candidate to base the hire or reject decision on.

Should we explicitly state during the interview that, although AI use on-the=job is encouraged, it is specifically 'forbidden' during the interview?

Yes, though I would make it a more generalized statement. For example: These tests are intended to evaluate your problem solving and programming skills, as such you are not to use outside tools to help solve them.

Is it ok for us to 'hide' AI detection 'tricks' in our exercises in order to try to sniff out AI use during interviews?

Yes. Trust but verify is important when interviewing. Just because someone lists 20 years of experience with a technology stack does not mean they actually have said experience. People do lie, but you do not know who, and so having ways to detect and catch deception before you hire such a person saves a lot of problems later.

Is it reasonable for us to reject this candidate for his use of AI during the interview?

Since you allow supplemental use of AI during normal work, and you hadn't expressly forbidden its use here it would probably be inappropriate to can this person's application purely on the basis of them having used it. On the flipside however, it would be perfectly appropriate if you wished to reject them on their inability to also provide a non-AI solution.

The thinking here being that they were capable on one aspect of the daily work but not another. Basically your interview process worked correctly to weed out candidates without the necessary skills for the role.

Should we explicitly state during the interview that, although AI use on-the-job is encouraged, it is specifically 'forbidden' during the interview?

For future applications it would probably be wise to state something to the effect that, while AI is used in some capacity at your employer, candidates are expected to be able to solve the tasks without AI assistance.

Is it ok for us to 'hide' AI detection 'tricks' in our exercises in order to try to sniff out AI use during interviews?

Yes, faking skills with AI is going to be this decade's version of lying on the resume, so detecting those who are trying to bluff their way through is totally valid.

We all feel that it should be obvious to a candidate that we're expecting them to solve our test exercises themselves without any outside assistance, but we do not explicitly state that during the interview.

Why depend on candidates thinking it is obvious? Why not be explicit? When interviewing candidates, I always state up front what I expect.

Is it reasonable for us to reject this candidate for his use of AI during the interview?

Yes, of course. You can reject the candidate for any reason you think indicates that they won't be the best candidate.

Should we explicitly state during the interview that, although AI use on-the-job is encouraged, it is specifically 'forbidden' during the interview?

If it's important to you, you should state this up front, so that there is no confusion. To me, it makes no sense to try and "trap" candidates.

You may also want to explain your thinking as to why this is important to you. Consider why you aren't instead testing for their ability to use AI effectively, since it may well be part of their everyday work.

Is it ok for us to 'hide' AI detection 'tricks' in our exercises in order to try to sniff out AI use during interviews?

Yes, it's okay to use tricks to detect "cheating". You can decide if you should warn the candidate about this detection method or not. I know I would tell them.

Remember - you are offering the opportunity, and interviews are conducted by your rules. You can require or prohibit anything that you think helps you in the hiring of good candidates.

That said, candidates are judging your company as well. You clearly will be weeding out candidates who think using AI during interviews is reasonable or even preferable - particularly when it will be used as part of the job.

Let me start by saying that I fall into the "AI is glorified autocomplete" camp and that I do not want to start a debate on what it is or is not.

I'm not going to start that debate, but instead I'll ask the question: Would you reject a candidate for using auto-completion during an interview?

Because for many developers (including some at your company by the sound it), AI tools are just as much a part of their workflow as auto-completion, or syntax highlighting, or all the other nice features that are built in to a modern IDE. Which means that there are plenty of developers out there who, when given a programming task, wouldn't think to ask "do I have to do this without CoPilot", any more than they'd ask "do I have to turn off auto-completion" or "do I have to do this on notepad without a proper IDE" or "do I have to do this without looking at the library documentation" or "do I have to do this with a pen and paper".

So if you want to impose arbitrary restrictions on how they're allowed to solve that task (which is a perfectly reasonable thing to do) then that's fine, but you have to clearly communicate those restrictions to the candidates.

Because it's not fair to penalise them because they didn't correctly guess which development aids you're happy for them to use and which you aren't, especially when the use of those tools is widespread in your company.

This is a fantastic Philosophical question

It comes down to this question: are you penalizing someone for using the tools that are available to them?

Now - if we take a quick trip down memory lane - it used to be considered taboo to admit to not knowing something and having to Google it in interviews.

There was a perception that in order to be in a particular position, one would have to know certain things, without resorting to Google.

That still holds true to some degree - if I was interviewing for a help-desk role for an IT-based company and someone didn't know what SMTP was, I might just disregard them. At the same time though, if I asked someone a fairly niche technical question and they were able to answer the broad strokes correctly and then say "I would have to Google that part", I would appreciate them being honest.

Furthermore - I tend to value people that know how to look for information they do not know, rather than people who might know that particular bit of information, but are incapable of research for things they don't know.

I said in my opening line that the question is whether you are penalizing someone for using the tools available to them.

As I write this answer it becomes increasingly clear that the deeper question is:

"Is this person capable of using the right tool for the job and interpreting the results?"

I, myself, do not really use AI. However, I do use Stack Exchange on occasion (no shit...) and as we all know, when we are coding or scripting, we run into a problem, we see some existing answers, copy and paste some code, tweak it to fit our use case - and then we continue.

The key requirement here can be thought of that the person doing it:

• understands the problem that they are trying to solve,
• can articulate the problem in such a way as to get help
• can evaluate the proposed solutions, both the pros and cons, and
• implements the solution required in accordance with best practices, standards etc.

I want you to now consider a series of potential candidates.

You give them all a series of problems.

One does it all without having to use any tools - I think we would all agree they really know their stuff and you would probably hire them on the spot.

One does it all using Google.

One does it all using Stack Exchange.

One does it all using AI.

I find myself leaning towards the idea that as long as the output meets the requirements, then the means by which they get there is less important. Whilst we might place a premium on the person that Googles and reads the documentation over the person who asks SE and over the person that uses AI - as long as the solution is implemented in a timely manner and conforms to your standards and does what you need it to do - I would view that as using the correct tool for the job.

And so to conclude

The reality is that any technical profession is going to be impacted by the use of AI - as such, I think the hiring process needs to account for that, as you would for using other tools.

To answer your questions directly:

Is it reasonable for us to reject this candidate for his use of AI during the interview?

It is only reasonable, in my mind, if they failed the key requirements - that is, they did not understand the problem or they implemented in such a way (e.g. blindly copying and pasting) that could cause problems.

Should we explicitly state during the interview that, although AI use on-the-job is encouraged, it is specifically 'forbidden' during the interview?

Whilst only you can make this decision, I think it is reasonable for some tasks - you state that the use of AI tools in that section is forbidden - but these should be scoped to the things that you consider to be 'core' concepts. For example, a networking engineer knowing some of the most common network ports and what they do (21, 25, 80, 443, 3389, 1433 etc.).

Is it ok for us to 'hide' AI detection 'tricks' in our exercises in order to try to sniff out AI use during interviews?

It's your circus and your monkeys. I think this would be unethical if you did not couple it with an expectation that a task be completed without the use of AI.

The key for making a hiring decision and an applicant using AI should be along the lines of "did they use AI to help them get a good answer, or did they use AI to cheat?"

You should try to treat "AI wrote this applicant's answer and now they try to pass it off as their own" the same as catching an applicant submitting code they copied from the internet (since that's pretty much what AI is, anyway) or getting a friend to write the answers for them and then turning them in. It's less an issue that the code was written specifically by AI, the real problem is that the code wasn't written by the applicant.

Likewise, an applicant asking an AI a question during an excercise should be treated the same as asking a colleague (or an interviewer). Look at the question, and ask yourself, would I consider it acceptable if they asked me directly? Am I okay with questions like these during a normal workday? Or is this person asking me to do their job for them?

This might be one of the few situations where you really can treat AI the same as a human being. If it's okay to ask you the question, it should be okay to ask an AI. If it's not okay to have another person do it, it's not okay to have an AI do it. If you suspect another person did the work, you can drill deeper and find out if that suspicion is correct, and you can attach consequences to finding out it is.

The question seems moot to me. It doesn't matter that the candidate used AI when you did not explicitly exclude AI use. Unless you asked and they specifically lied and said that they didn't use AI (then potentially a character flaw you might not want to have in your team).

Given that OP said

In this specific case, the answer to "Is this person capable of using the right tool for the job and interpreting the results?" is, unfortunately, "No". Had this candidate interpreted the results of his AI query instead of blindly copying them then we'd probably never have known that he used AI to solve the exercise.

the actual issue is that the candidate did not solve the problem correctly. That is a good reason to reject the candidate. If they had solved it by using AI and then adjusted the code to the problem (or adjusting the prompt so AI gets it right), it would have been a good fit, given you do use AI in your daily work.

It is still a good idea to make the rules clear in advance (either allow using AI or not) to avoid this detraction from your actual test goals and any uncertainty in the candidates.

Is it reasonable for us to reject this candidate for his use of AI during the interview?

Yes.

I have personally called out a candidate for using AI to answer technical questions during an interview, and then ended the interview. I don't have time to waste on someone who needs a crutch to answer basic programming questions.

AI should be used as a tool. It should not be a crutch.

If a programmer needs to use AI to answer simple technical questions and/or solve a simple problem, they are not competent.

A person who has delegated their problem solving skills to a machine is not going to be able to solve complicated problems. Such as debugging a C program.

In my opinion the only thing you might blame the applicant for is that they did not ask whether AI usage is allowed for the test.

Even though it was not explicitly forbidden it would be a sensible thing to ask for permission.

Personally I do think that AI is a valid tool and coming up with a solution through the use of AI is no more “cheating” than using an algorithm or design pattern you did not invent on your own.

Is it reasonable for us to reject this candidate for his use of AI during the interview?

No, not hiring a candidate for using a tool that your development environment uses and promotes is the very definition of being unreasonable. Why would you not allow a prospective candidate to use the tools you use yourself?

Should we explicitly state during the interview that, although AI use on the job is encouraged, it is specifically 'forbidden' during the interview?

Although this would promote transparency, it may also reduce the applications, as many would wonder why anyone would have such a rule.

Is it ok for us to 'hide' AI detection 'tricks' in our exercises to try to sniff out AI use during interviews?

What you should do is remind yourself that this is a coding interview and you are not a government operative in a Robert Ludlum spy thriller. That is what you should do

If you tell them before the test “don’t use any AI tools, we want to know what you can do, not some AI tool” then you can count any task solved using AI as “failed”. If you didn’t tell them, then you can count everything handled incorrectly by AI as failed. And everything they cannot explain, because I expect you to explain any solution you gave. You can also tell them “sorry, we didn’t tell you about our no-AI rule, please repeat the test without using AI”.

It is reasonable, yes, though if it's done covertly then it might be a reason to select a different candidate.

Contacting all the candidates to ask them about their approach to AI could give useful insight.

If you don't want candidates to use AI because you only want to test innate ability, then explicitly state that next time you are interviewing.

Don't use tricks to try to trap candidates; ask them to share their entire desktop screen on the call, provide a sandboxed environment, or interview on premises, if you want exam-like conditions.

Going forward, you could ask candidates about their attitude to AI as part of the interview, before doing the tests, which might give useful insight.

The candidate did not ask if it was ok to use AI assistance, and typed the code to solve the exercise as if he had thought it up himself.

How could you possibly come to that conclusion by how they typed the code? Additionally, while they didn't explicitly ask, you also didn't explicitly forbid it. It seems to me that a smart candidate would use any and all resources available to them. Including AI and search engines.

We all feel that it should be obvious to a candidate that we're expecting them to solve our test exercises themselves without any outside assistance, but we do not explicitly state that during the interview.

Why would it be obvious? Unless you explicitly address it then it isn't obvious. Don't assume what people know or assume about anything.

TBH, my personal opinion is that I'd be looking for candidates that exhibit a willingness to go outside of their own knowledge and experience to solve problems.

If a candidate uses AI during an interview in a way that feels off get them to explain aspects of their CV over a phone call. Something simple like, “Can you talk me through how you’d actually handle that?” usually reveals if they really know their stuff or not.

I think it's important to realise that Chat GPT can be used legitimately like practicing or using for editing or even coming up with ideas. But using it to develop a whole application or to anwser interview questions crosses a line. I think AI will become how everything is done in the future. If they're putting no effort in and just getting AI to write the whole thing that's a poor quality application and should be rejected. But if the application looks good and it's had input from AI that's just the new normal.

Afterward, make a note of it and discuss with your team. Be consistent — if it felt like an honest mistake, maybe give them another shot. If it felt like they were gaming the process, move on.

Going forward, it might help to include a quick heads-up in your interview invite: prep with AI if you want, but we expect real answers during the call. At the end of the day, it’s not about the tool — it’s about whether the person behind it can actually do the job.