Job Interviews Are Broken (www.theatlantic.com)

TikTok clips and new “interview copilot” apps have exposed a fast-moving arms race: employers use AI to screen résumés and run automated first-round interviews, and candidates are fighting back with generative-AI assistants that whisper answers, craft bespoke résumés, or even listen in real time during Zoom calls. Vendors such as Final Round AI (which offers role templates, live suggestions, and a paid tier that hides the app during screen share), LockedIn AI, and influencer-promoted tools promise to help candidates “crush” interviews—mostly for software roles—and TikTok creators both demonstrate and monetize these practices. Some viral posts are staged marketing, but real products exist and produce plausible, formulaic LLM responses that could get people past ritualized interviews without proving on-the-job competency. For the AI/ML community this matters technically and ethically: it creates a detection/evadation problem (real-time LLM assistance vs. liveness/screen-share detection), undermines signal quality in training data used for hiring models, and amplifies fairness concerns when automated filters reward surface polish over genuine skill. Employers are responding (some urging in-person interviews), but the deeper issue is systemic—automation has degraded hiring pipelines into gatekeeping rituals, incentivizing applicants to “pass” rather than demonstrate competence. Designers of hiring systems must therefore balance robustness (behavioral/liveness signals, task-based assessments) with fairness and transparency to avoid an escalating cat-and-mouse game driven by generative models.