Google's AI team has 0% faith in HR filters — a hiring wake-up call

For HR leaders evaluating AI screening tools, the news that Google's own AI researchers are circumventing their company's filters should be a red flag. If the creators of advanced AI don’t trust their systems to evaluate resumes fairly, why should your organization? This incident exposes the critical gap between AI hype and HR reality.

Google, a company that sells AI-driven hiring tools to enterprises, is now at the center of a credibility crisis: its own elite AI researchers have publicly signaled that those tools cannot be trusted. A document created by Google DeepMind's AGI Safety and Alignment Team, obtained by Bloomberg, encourages job applicants for the team to fill out a special form to bypass the company's internal AI screening. The reason stated is a 'non-trivial probability that your CV will be screened out incorrectly or take too long to reach us.' The document includes the warning 'PLEASE DO NOT SHARE THIS DOC WIDELY,' highlighting the sensitivity of the team's lack of confidence. This revelation lands at a particularly awkward moment. Google Workspace, the suite that includes Gmail, Docs, and Drive, now aggressively markets AI features to HR departments, promising to 'save HR time by quickly creating drafts for job postings, evaluating resumes, and forecasting hiring needs.' The disconnect between internal practice and external marketing could not be starker: if the very researchers building Google's most advanced AI don't trust the system to handle resumes, why should any corporate customer?

“

A document created by Google DeepMind's AGI Safety and Alignment Team, obtained by Bloomberg, encourages job applicants for the team to fill out a special form to bypass the company's internal AI screening.

The incident also brings fresh attention to broader industry concerns about AI in hiring. A Bloomberg investigation earlier in 2026 uncovered discriminatory outcomes in AI hiring systems, with some models penalizing candidates based on gender or ethnicity. Google's own AI Principles commit to avoiding unfair bias, but the escape hatch created by its safety team suggests that at least one group within the company believes the systems are not yet reliable enough for high-stakes selection. Google DeepMind's spokesperson denied that the filters incorrectly screen applicants, stating that the form is simply 'to go past the recruiter review, and get their resumes direct to the people on the team.' Yet the need for a backchannel and the 'non-trivial probability' language imply a deeper problem, one that could have legal and regulatory ramifications. The European Union's AI Act classifies AI systems used in employment as 'high-risk,' requiring rigorous conformity assessments and human oversight. Google's internal dissent may accelerate calls for stricter enforcement of such laws, and could prompt companies using similar tools to demand transparency audits.