Google Admits AI Algorithms Often Reject Worthy Resumes: Candidates Will Be Selected by Real People

The AI safety team at Google DeepMind, known as the AGI Safety and Alignment Team, has advised candidates to submit resumes through a special form so that applications are definitely seen by human employees, as reported by Bloomberg. The developers fear that the AI hiring system may mistakenly reject resumes or pass them on too late. Google itself, however, offers corporate clients AI features of its services for resume screening and other hiring tasks. In a document reviewed by Bloomberg, the developers explain that the AI system could erroneously discard resumes or delay them, and filling out the form guarantees that a real person from the team will see the application. The authors also ask not to widely distribute information about the new form. Google DeepMind told Bloomberg that the separate form is not meant to bypass AI errors but to ensure resumes reach team members directly, bypassing recruiter screening. Google stressed that the form does not guarantee employment and does not replace the standard hiring process. As Bloomberg notes, Google itself offers businesses AI solutions for HR tasks such as preparing job postings, resume screening, and forecasting staffing needs. However, AI systems in the market are criticized for possible erroneous decisions and discrimination. Previously, Bloomberg found signs of potential bias in ChatGPT from OpenAI when evaluating candidates, and Workday faced similar accusations that its AI hiring systems might consider race, age, and disability; Workday denies the claims and says final decisions are made by humans. While candidates use neural networks to pass automatic filters, Google DeepMind asks candidates to avoid AI when filling out resumes because human staff will read applications, and AI-generated texts often sound very similar. The document states that HR specialists are very tired of reading LLM-written responses because they all sound alike.