Recruiters deploy AI to catch ghost coders, AI-assisted cheating

Bengaluru: The rise of GenAI is exposing new vulnerabilities in campus hiring, as recruiters confront increasingly sophisticated cheating techniques—from hidden wearable devices and AI-assisted tools to proxy candidates and ghost coders taking control of assessments through remote desktop software.Recruiters say organised “interview-as-a-service” networks are making hiring fraud more sophisticated, prompting companies to deploy AI-powered proctoring as a key line of defence. Many Indian tech firms use AI-powered proctoring platforms like Talview and Mercer Mettl and assessment platforms like Hackerearth and Hackerrank to safeguard the integrity of hiring assessments.According to Vivek Ravisankar, co-founder and CEO of tech hiring platform HackerRank, hiring integrity is being undermined on three fronts: leaked assessment questions, candidates using AI-assisted tools such as Cluely, Interviewman and remote-desktop applications during tests, and outright impersonation, where another individual takes an assessment or interview on behalf of the candidate.“The issue is quite widespread, especially in university recruiting in India. We flag about 30%-35% of sessions with at least one suspicious behaviour. The biggest issue is the use of AI-powered cheating apps,” Ravisankar told TOI.To combat this, companies are moving away from browser-based assessments. AI-powered proctoring analyses webcam feeds, audio and screen activity to detect suspicious behaviour and safeguard the integrity of large-scale assessments.

More than half of HackerRank’s university hiring assessments now run on a secure desktop application with candidates’ cameras switched on. The platform automatically blocks known cheating applications, disables unauthorised software and creates a controlled testing environment.HackerEarth says cheating methods have evolved just as rapidly. According to its CEO, Vikas Aditya, the four most common forms of malpractice today are AI-generated code submissions, proxy candidates hired through Discord or Telegram groups, off-camera assistance through secondary devices or another individual feeding answers, and virtual machines or remote-desktop software that conceal a second AI session from proctoring systems.“The common thread is that nearly all of these exploit the same weakness—a static assessment that scores only the final answer without observing how it was produced,” Aditya said. “Our most effective defence is a short live follow-up interview where candidates explain their solution. Most candidates who relied on AI fail within two questions.”The shift is already visible in employer behaviour. HackerEarth’s 2025 Technical Hiring Landscape Report shows the share of companies using proctored technical assessments rose from 64% at the beginning of 2025 to a peak of 77% by July.

Across the year, nearly two-thirds of all technical assessments were proctored, reflecting growing concern over AI-enabled cheating.The industry believes many companies still underestimate the scale of the problem because conventional coding tests reveal only the final score—not how it was achieved.“It helps to separate assistance from impersonation,” Aditya said. “One is a genuine candidate using AI. The other is somebody else taking the assessment through proxy interviewers, deepfakes or remote-desktop ‘ghost coders’.

These require very different countermeasures.”Companies are therefore adopting layered defences. Secure browsers block copy-paste, multiple tabs and external applications, while identity verification combines government ID checks, facial matching and liveness detection to prevent proxy candidates. AI-based plagiarism detection and webcam monitoring help identify AI-generated responses, hidden devices and the presence of another person.Yet hiring platforms acknowledge the battle is becoming an arms race. “No single defence is foolproof,” Aditya said. “The most durable solution is changing what we measure. AI can generate answers, but it struggles to defend reasoning in a dynamic conversation. That’s why companies are increasingly testing problem-solving ability instead of simply checking whether the final answer is correct.”