Half of AI Job Candidates Cheat in Interviews, Report Finds.

The Scale of Demand

AI companies are hiring at a pace that outstrips the supply of verified talent, and the compensation reflects it. Zero G Talent's board data maps the salary landscape across two of the sector's most aggressive recruiters:

Beyond the bands above, Anthropic's individual postings — Performance Engineer for inference engines, Research Engineer for chip design reinforcement learning, and Staff+ Research Engineer for RL data platforms — reach $850,000 per year, while Databricks's Senior Director of Enterprise for Retail Vertical Strategic Accounts runs $440,000–$605,000 and Director-level Lakebase Sales Specialist roles land between $430,400 and $591,800. Both companies added dozens of new postings in a single week, competing for a narrow pool of candidates who can prove they will deliver at the top of their field.

The volume is high. The bar is not.

The Screen

Taiga — the Centre for Transdisciplinary AI at Umeå University in Sweden — organized its TAIGA Days event for November 19–21, built around four strategic themes: Arctic AI, AI and Education, Power and AI, and AI for the region. The center's stated goal is "to foster AI research and collaboration across disciplines and sectors for the benefit of society." Each theme received a half-day session, registration closed November 16, and participation was free. The four themes reveal what the field treats as its most pressing questions: Arctic AI targets AI theory and systems tied to Arctic environments; AI and Education aims to create roadmaps for AI curricula; Power and AI examines how AI reshapes societal power balances; and AI for the region reaches out to connect AI research at university with needs of municipalities, companies and other regional organizations.

But the hiring side of the AI industry operates on a different logic. Companies filling roles at the top of the pay scale have moved well past résumé review. The first filter is a practical exercise: debugging a live system, navigating a distributed training pipeline, or reasoning through a production failure under time pressure. This is not a trivia quiz; it is a direct simulation of the work, designed to surface whether a candidate can operate at the level the role demands from day one.

After the practical exercise, candidates face a multi-stage interview loop. At least one technical deep-dive with a senior engineer probes not just whether the candidate knows a solution but how they arrived at it, including the reasoning, the trade-offs, and the dead ends explored. A second stage typically involves a cross-functional conversation, where a candidate's ability to communicate technical decisions to non-technical stakeholders gets tested. A third layer, when applicable, is a portfolio or project review that examines past work for depth, not breadth. The process does not reward candidates who have packaged a dozen shallow projects; it rewards those who can walk through a single complex piece of work and explain the engineering choices behind it.

The rigor is not arbitrary. AI companies operate in a market where the cost of a bad hire runs to the mid-hundreds of thousands in salary alone, not counting lost productivity and team disruption. A screening process that takes longer and demands more from each candidate is a practical response to a hiring market where mistakes cost millions. The consequence is that candidates who relied on polished résumés and brand-name affiliations now face rejection at every stage. The screen does not care where you went to school or which logos appeared on your previous title. It cares whether you can solve the problem in front of you, under observation, with clarity and rigor.

The Industry Shift

Taiga's approach mirrors a broader shift across the AI industry, where companies are moving away from resume-based evaluation toward skills-based assessment, a transition that has been building for years and is now accelerating under competitive pressure.

A 2024 Workday survey of 2,300 business leaders found that 55 percent had already begun transitioning to a skills-based talent model, with most of the rest planning to start within 12 months and only 2 percent of HR leaders reporting successful adoption across all their processes. That figure suggests momentum, but the execution gap is enormous. Deloitte's research found that nearly 40 percent of skills required on the job are set to change, and most employers cite the skills gap as the key barrier to business transformation, a finding echoed by 60 percent of surveyed businesses and 72 percent of CEOs who report that talent gaps lead to critical business challenges.

The AI sector sits at the center of this tension. The Brookings Institution documented that between 2000 and 2019, a paper ceiling of degree screens, biased algorithms, and stereotyping cost STARs — skilled workers through alternative routes — access to almost 7.5 million jobs, a figure dwarfed by today's 70 million-strong STAR workforce, more than 30 million of whom have demonstrated skills qualifying them for higher-wage positions. Yet traditional hiring filters continue to exclude them. AI companies, which depend on talent that may have acquired expertise through bootcamps, open-source contributions, or self-directed projects rather than four-year degrees, face a particular incentive to look past paper credentials.

Deloitte documented cases where skills-based hiring cut time-to-fill from 127 days to 47, boosted internal mobility by 45 percent, and delivered a 340 percent return on investment within two years. But the Brookings Institution also cautioned that the same algorithms and automated systems companies deployed to sort talent have, at times, reinforced the paper ceiling rather than tearing it down. "Employers changed core business practices and began to rely on flawed algorithms to sort through applications and identify and evaluate talent," the institution said. The lesson is that raising the bar requires intentionality, a skills-based screen that is genuinely applied, not one that simply replaces a degree requirement with a different hidden filter.

Cheating, Caught

Candidates trying to land roles at AI companies face a screening process transformed by the very technology these companies build. The contest between candidates seeking to impress and the tools available to game the system has become central to how AI firms evaluate talent.

A CNBC investigation published in March 2025 found that more than 50 percent of candidates cheated during technical interviews, using tools designed to circumvent webcam-based proctoring. One such tool, Interview Coder, was marketed as "webcam-proof" by its creator, Lee, whose website claimed the product was immune to screen detection features available on Zoom and Google Meet. Lee said the company was on track to hit $1 million in annual recurring revenue by mid-May, a figure that underscores how lucrative the industry around tools that help candidates cheat had become.

Companies have responded with a multi-layered counter-strategy. Anthropic, the maker of Claude, issued new guidance in February requiring candidates not to use AI assistants during the hiring process. Margaret Callahan, an Amazon spokesperson, said the company asks candidates to acknowledge they will not use unauthorized tools during interviews or assessments. When Amazon discovered that a candidate named Lee had used such tools, the company rescinded offers that had been made, a consequence that signals the stakes are real, not theoretical.

The broader industry is also rethinking the format of evaluation itself. Deloitte reinstated in-person interviews for its U.K. graduate program, per a September report. Google CEO Sundar Pichai said during a February town hall that hiring managers should consider returning to in-person interviews. These moves reflect a recognition, as CNBC reported, that "the combination of rapid advancements in AI, mass layoffs of software developers, and a continuing world of remote and hybrid work has created a new problem for recruiters."

Kirk, founder of a startup in this space, said his company was considering moving to in-person interviews, though he acknowledged that potentially limits the talent pool. Meanwhile, Leetcode Wizard's De Vries said his product's goal was to make "leetcode interviews a thing of the past," framing the entire traditional screening apparatus as an artifact worth dismantling. The tension is clear: candidates who invested in gaming the old system now face a market that is actively dismantling it.

The practical takeaway is not a hack but a pivot. Build skills that survive scrutiny without technological aid. The resume may open the door, but it is the unassisted demonstration of competence that gets you through it.

Who Gets Left Behind

The shift toward skills-based screening does not happen in a vacuum. Nine in ten U.S. employers now use AI screening tools to sort and rank job seekers, with most relying on the same few third-party vendors, per Stanford HAI. Among Fortune 500 companies, that figure reaches 98.4 percent, per Brookings research. When companies tighten their own screening to require deeper practical demonstrations, they join a system that already filters out millions of applicants before a human ever reviews a resume.

Stanford HAI researchers found that 26 percent of Black applicants and 15 percent of Asian applicants applied to positions where the AI system discriminated against their racial group. If the system had recommended Black and Asian candidates at the same rate as the most-favored group — typically white applicants — roughly 40,000 more applications would have advanced to the next stage of hiring. One in ten applicants who submit four applications faces rejection from every single place they apply.

Gender bias compounds the problem. A Brookings study found that men's-name resumes were favored 51.9 percent of the time, while women's names were favored just 11.1 percent. Racial bias proved even starker: white-associated names won preference in 85.1 percent of tests, while Black-associated names led in only 8.6 percent. Resumes with Black women's names appeared 0 percent of the time compared to white men's names. Amazon's own experience reinforces the pattern: in 2018, the company revealed that an AI recruiting tool it developed unfairly discriminated against graduates of all-women's colleges, suggesting that educational history can be used to infer and discriminate against particular identities. These patterns predate any company's current process, but they define the pool from which any skills-based screen must now select.

The entry-level market bears a particular burden. Stanford's Digital Economy Lab documented that within firms, entry-level hiring in AI-exposed jobs declined 13 percent relative to less-exposed roles, with these impacts appearing only after the proliferation of large language models. Employment declines concentrated among 22- to 25-year-old workers in software development, customer service, and clerical work. At the same time, companies see nearly three times as many applications for entry-level positions as in 2022, per Stanford HAI, a flood that AI screeners manage by narrowing the funnel before human review begins.

Regulation is scrambling to catch up. New York City's AI hiring law took effect in 2023, but Brookings researchers identified weaknesses that have limited its ability to meaningfully reduce discrimination. Colorado's AI Act, set to take effect in 2026, mandates auditing of AI hiring systems and gives applicants the right to appeal adverse algorithmic decisions. California became the first state, in September 2024, to officially recognize intersectionality as a protected identity, while Texas enacted AI nondiscrimination provisions, though Brookings noted that developers and deployers there are liable only if they intended to.

Anthropic and Databricks are adding dozens of roles each week. The screens that filter who fills them will only get tighter, and the people most likely to be screened out are the ones the industry says it cannot find.

Working in AI? Zero G Talent tracks the openings: see every open Databricks role, browse AI jobs, openings at Anthropic, and the people building the field.