Google DeepMind Safety Team Bypasses Company's Own AI Recruitment Tool
Google DeepMind's AGI safety and alignment team reportedly bypassed Google's internal AI recruitment system by asking candidates to submit their resumes via a separate form. The team expressed concerns that the automated AI system might mistakenly filter out qualified applicants or delay their applications. This incident highlights a significant irony and lack of trust in AI recruitment tools, even among the very AI researchers developing advanced systems. It underscores the ongoing challenges and skepticism surrounding the reliability and fairness of automated hiring algorithms in the tech industry. An internal document warned candidates that Google's application system had a "not insignificant probability" of filtering out their resumes. Although a DeepMind spokesperson claimed the internal system does not erroneously eliminate candidates, they confirmed the team used a separate form to bypass initial recruiter screening.
## BACKGROUND
AI recruitment tools are increasingly used by companies to screen resumes, rank candidates, and save HR time, but they have faced criticism for bias and inaccuracy. For instance, historical data used to train these models can lead to systemic rejection or reproduce diversity biases, as seen in past cases like Amazon's scrapped hiring tool.