Which companies ban AI in job interviews in 2026?

Among firms with a written rule: Hudson River Trading, Two Sigma, IMC, Akuna Capital (on its summer 2027 software and platform intern postings), Goldman Sachs, Anthropic and Microsoft, plus Amazon for any tool it has not provided. Jane Street, Jump, DRW, Optiver US, Five Rings and Point72 had no AI clause in any posting on 28 October 2026, but every written policy above defaults to no AI unless you are told otherwise, so treat silence the same way.

Can you use AI in an Amazon interview?

Not your own. Amazon's internal guidance, reported by Business Insider in February 2025, asks candidates not to use GenAI during interviews unless explicitly permitted and warns of disqualification. Since January 2026 Amazon's software engineering assessment includes Amazon's own interactive AI assistant (aboutamazon.com), and its SDE Early Career 2027 posting lists experience with AI-assisted development tools as a basic qualification.

How do companies detect AI use in online assessments?

Mostly through the assessment platform. HackerRank's Proctor Mode takes screenshots every 15 seconds (every 5 after a window switch), watches the webcam for extra faces and phones, flags a second monitor and blocks pasting. Its July 2026 release added gaze detection in limited availability. CodeSignal says the most common signals in its flagged 2025 sessions were off-screen referencing (35%) and unusually linear typing (23%).

Can HRT rescind an offer if it finds you used AI?

Its postings say so. All 86 Hudson River Trading postings on Greenhouse on 28 October 2026 said AI use in interviews or assessments is strictly prohibited unless agreed, and that HRT may end the interview, disqualify the candidate and/or rescind any job offer already extended.

Is using AI on my resume or application banned too?

Usually not. Anthropic's candidate guidance (July 2025) asks you to write the first draft yourself and then lets you use Claude to refine it, but not to invent experience. Microsoft's candidate code of conduct encourages responsible AI use while preparing. The bans in this list cover assessments and interviews.

Do in-person interviews stop AI cheating?

Employers are betting they help. Sundar Pichai said in June 2025 that Google would add at least one in-person round for engineers. Checkr's October 2026 survey of 3,000 US hiring managers found real-time video interviews dropped from 55% to 42% as rounds moved in person, and 60% said AI made virtual interviews harder to trust.

Blog28 Oct 202611 min readApplr Team

Companies That Ban AI in Interviews (2026) and How They Catch It

HRT can rescind an offer, Two Sigma acts on perceived use, Amazon and Meta hand you their own AI. The firm-by-firm list for 2026 and how assessments flag you.

Most firms that hire new grads now ban AI in the interview and ask for AI skills in the job ad. In 2026 the rule depends on who supplies the AI. Quant firms and banks ban it outright (HRT reserves the right to rescind an offer). Big Tech bans your tools but increasingly hands you its own: Amazon's assessment assistant, Meta's AI round, Google's Gemini pilot. Only a few companies, Canva among them, require it. And detection does not need proof: proctoring software logs your eyes, screen and paste events, and Two Sigma says perceived use is enough to disqualify you.

Below: the firm-by-firm list from primary sources, the three kinds of ban, how detection works in an assessment and in a live round, the habits that get honest candidates flagged, and what to do this week.

The 2026 list: who bans, who allows, who requires

Company	Rule on candidate AI	Stated consequence	Source
Hudson River Trading	Banned in interviews and assessments unless agreed	End the interview, disqualify, and/or rescind an offer	All 86 Greenhouse postings, 28 Oct 2026
Two Sigma	ChatGPT or any other AI tool strictly prohibited	Perceived use can disqualify, at Two Sigma's sole discretion	SWE interviewing page
IMC	No AI, code generators or forums unless permitted; recording banned too	May disqualify	Interview code of conduct
Akuna Capital	AI not permitted in interviews or assessments	Disqualification	4 of 34 postings (summer 2027 software and platform interns)
Goldman Sachs	No external sources, ChatGPT or Google search included	None stated	Campus email, Fortune, Jun 2025
Anthropic	You draft, Claude can polish; take-homes and live rounds AI-free unless told	None stated	Candidate AI guidance, Jul 2025
Microsoft	AI fine for prep; no outside help in assessments or interviews unless permitted	None stated	Candidate code of conduct
Amazon	No GenAI in interviews unless permitted; its own assistant in the SDE assessment	Disqualification	Business Insider, Feb 2025; Amazon, Jan 2026
Meta	Classic rounds AI-free; one onsite round has a built-in assistant	Not stated	Meta prep guide; Business Insider, Jul 2025
Google	At least one in-person round; Gemini allowed in a piloted round from H2 2026	Not stated	Pichai, Jun 2025; Business Insider, May 2026
Canva	AI expected (Copilot, Cursor, Claude) in backend, ML and frontend rounds	Not applicable	Canva engineering blog, Jun 2025
Jane Street, Jump, DRW, Optiver US, Five Rings, Point72	No AI clause in any posting	Not stated	Greenhouse feeds, 28 Oct 2026

Silence in a posting is not permission. Every written rule above bans AI unless the company says otherwise, so treat any round without an explicit invitation as AI-free.

The same firms want AI in the job itself. Akuna's ban opens by calling Akuna an AI-friendly company that encourages employees to use AI daily. Seven of Jump's campus internship postings list hands-on use of AI tools such as Claude, ChatGPT or Claude Code as a requirement or a plus, and Jane Street's compiler engineer postings ask for someone willing to use AI assistance to ship faster. Amazon's SDE Early Career 2027 posting (19 October 2026) lists experience with AI-assisted development tools as a basic qualification. Even Cognition, which builds an AI coding agent, does not allow AI in interviews; its people lead compared it to a maths test without a calculator (Business Insider via Moneywise, May 2026). The logic is consistent: the interview measures what you can do unaided, so the firm can trust what you do with AI later.

Three kinds of ban, and why the wording matters

1. The ban with an offer clawback. HRT's clause, on every one of its 86 live postings (all updated between 1 and 26 October 2026), goes past the interview: if HRT determines AI was used, it can end the session, disqualify you, and/or "rescind any job offers extended". It is the only policy in this list that names what happens after an offer.

2. The ban on suspicion. Two Sigma's software engineering page says "perceived usage" of AI could disqualify you, and that Two Sigma decides in its sole discretion. There is no standard of evidence to argue with. In practice, looking like you are using AI counts the same as using it.

3. The ban "unless we say so". Anthropic, Microsoft, IMC and Amazon all carve out an exception, and that exception is where the sanctioned AI rounds live. Amazon is the clearest case. In February 2025 Business Insider reported internal guidance asking candidates not to use GenAI during interviews unless explicitly permitted, on the grounds that it gives an unfair advantage. In January 2026 Amazon said it had rebuilt its software engineering assessment so candidates work with an interactive AI assistant on a real coding platform, and that 87% of candidates said the questions related to the role. The assistant Amazon gives you is permitted; the one on your phone is not. Meta works the same way: its prep guide still lists tools like ChatGPT as unauthorized help, while one onsite round runs with an assistant built into CoderPad (our breakdown of that round).

How they detect it

Detection is layered: the assessment platform, the live interviewer, then identity checks and in-person rounds.

Layer	What is watched	Source
Online assessment (HackerRank Proctor Mode)	Screenshots every 15 s, every 5 s after a window switch; webcam presence and extra faces; phones or tablets in frame; a newly connected monitor; tab and full-screen exits; pasting disabled; result graded High or Medium with session replay	HackerRank help center
New in 2026 (HackerRank)	Gaze detection (looking away, then resuming typing; medium severity, limited availability); screenshot analysis tuned to newer AI tools; its AI interviewer pauses if you add a monitor or share your screen	HackerRank July 2026 release notes
Online assessment (CodeSignal)	Of flagged 2025 sessions: 35% frequent off-screen referencing, 23% unusually linear typing, 15% similarity to known or leaked answers	CodeSignal, Feb 2026
Live interviewer	Perfect code with no iteration; an optimal answer with no reasoning; explanations that don't match the code; large blocks appearing at once; no feel for edge cases	Karat, Mar 2026
Identity and in-person	Government ID checks (61% of managers); in-person rounds replacing video	Checkr, Oct 2026

The pressure is highest at entry level. CodeSignal says the flagged attempt rate on its proctored assessments more than doubled in 2025, from 16% to 35%, and on early-career assessments it rose from 15% to 40%, with Asia-Pacific at 48% against 27% in North America. These are a vendor's figures for flagged attempts, not confirmed cheating, but they explain why new-grad tests are the most locked down.

Live rounds are moving off video. Checkr surveyed 3,000 US hiring managers from 11 to 16 September 2026 (published 15 October). It found 67% had suspected a candidate of using AI to misrepresent themselves, up from 59% in 2025; 40% had interviewed someone later revealed to be using a fake identity; and 34% had confirmed a proxy interviewee. Real-time video interviews fell from 55% to 42%. Google moved earlier: Sundar Pichai said on the Lex Fridman Podcast in June 2025 that engineers would get at least one in-person round to check fundamentals (HCAMag, Aug 2025).

Interviewers mostly catch it by asking questions. In a 2025 Blind thread on getting caught, a new grad wrote in December 2025 that an AI overlay never showed on the shared screen, but the interviewer noticed the code was far more polished than the resume suggested, and the candidate was not advanced. In the same thread an Amazon-tagged interviewer said they ask candidates to show their hands on camera. In an older Blind post (March 2022), an interviewer spotted a candidate glancing at a second monitor and speaking as if reading. Google- and Amazon-tagged replies said they would recommend no hire or blacklist the candidate.

What gets honest candidates flagged

Writing code elsewhere and pasting it in. Proctor Mode blocks large pastes and logs the attempt, and CodeSignal counts unusually linear typing as a signal. Type in the platform's editor, mistakes included.
Looking away to think. Off-screen referencing is the most common signal CodeSignal flagged, and HackerRank's gaze detection looks for exactly that: looking away, then typing. If you need to think, keep your eyes on the problem, or tell the interviewer before you sketch on paper.
Notes in another language. HackerRank's conversation detection watches for repeated typing and deleting that looks like messaging, including regional languages typed in Latin letters. Pinyin notes typed into the editor fit that pattern. Write your notes as English comments and leave them in.
A second screen or device nearby. New monitor connections are flagged, and so are phones and tablets visible on the webcam. Unplug and clear the desk before you start.
A clean answer you can't walk through. Every interviewer account above turns on the gap between code and explanation. Practise narrating your choices, not just reaching the answer.
Misusing a permitted assistant. In Meta's AI round and Amazon's assessment the AI is allowed, but pasting output you cannot defend still fails.

You will not see your integrity report. HackerRank tells reviewers to check the session replay and notes that many customers ignore Medium flags, but Two Sigma's standard is perception, and HRT's is whatever it determines.

What to do this week

Read the clause. Check the posting and the invite email for an AI rule and screenshot it. If there is none, email the recruiter and ask which rounds allow AI and which tool.
Practise AI-free, out loud. Timed sessions where you narrate tradeoffs and edge cases. Applr's company mock interviews run with no assistant on your side, which is the format the banning firms test: try the HRT quant researcher mock or the Two Sigma software engineer mock.
Set up the desk. One monitor, phone out of frame, meeting assistants and overlays closed, notes in English.
Train the permitted-AI format separately. For Amazon, practise checking and explaining code an assistant hands you; the Amazon SDE1 mock covers the rest of the loop.
Watch the deadlines. Our New Grad 2027 tracker lists which 2027 programs are open now.

Practice

Sources

Frequently asked questions

Which companies ban AI in job interviews in 2026?: Among firms with a written rule: Hudson River Trading, Two Sigma, IMC, Akuna Capital (on its summer 2027 software and platform intern postings), Goldman Sachs, Anthropic and Microsoft, plus Amazon for any tool it has not provided. Jane Street, Jump, DRW, Optiver US, Five Rings and Point72 had no AI clause in any posting on 28 October 2026, but every written policy above defaults to no AI unless you are told otherwise, so treat silence the same way.
Can you use AI in an Amazon interview?: Not your own. Amazon's internal guidance, reported by Business Insider in February 2025, asks candidates not to use GenAI during interviews unless explicitly permitted and warns of disqualification. Since January 2026 Amazon's software engineering assessment includes Amazon's own interactive AI assistant (aboutamazon.com), and its SDE Early Career 2027 posting lists experience with AI-assisted development tools as a basic qualification.
How do companies detect AI use in online assessments?: Mostly through the assessment platform. HackerRank's Proctor Mode takes screenshots every 15 seconds (every 5 after a window switch), watches the webcam for extra faces and phones, flags a second monitor and blocks pasting. Its July 2026 release added gaze detection in limited availability. CodeSignal says the most common signals in its flagged 2025 sessions were off-screen referencing (35%) and unusually linear typing (23%).
Can HRT rescind an offer if it finds you used AI?: Its postings say so. All 86 Hudson River Trading postings on Greenhouse on 28 October 2026 said AI use in interviews or assessments is strictly prohibited unless agreed, and that HRT may end the interview, disqualify the candidate and/or rescind any job offer already extended.
Is using AI on my resume or application banned too?: Usually not. Anthropic's candidate guidance (July 2025) asks you to write the first draft yourself and then lets you use Claude to refine it, but not to invent experience. Microsoft's candidate code of conduct encourages responsible AI use while preparing. The bans in this list cover assessments and interviews.
Do in-person interviews stop AI cheating?: Employers are betting they help. Sundar Pichai said in June 2025 that Google would add at least one in-person round for engineers. Checkr's October 2026 survey of 3,000 US hiring managers found real-time video interviews dropped from 55% to 42% as rounds moved in person, and 60% said AI made virtual interviews harder to trust.

Practice for top companies

Browse all interview guides View all articles