Alphabet Inc. (GOOGL) markets its artificial intelligence tools to corporate clients as a way to streamline recruitment, promising to help human resources departments sift through mountains of resumes and identify top talent faster. But inside one of the company’s most advanced AI research units, staffers are taking a decidedly different approach: they are warning job applicants that Google’s own automated screening systems might wrongly toss out their applications.
The AGI Safety and Alignment Team at Google DeepMind, the division focused on mitigating the risks of highly capable AI, has set up a separate form for candidates applying to its open positions. The document, viewed by Bloomberg and marked with the disclaimer “PLEASE DO NOT SHARE THIS DOC WIDELY,” tells prospective hires that the standard application pipeline carries a “non-trivial probability” of incorrectly screening out their CVs or delaying their delivery to the team.
“We have an applications system with a non-trivial probability your CV will be screened out incorrectly or take too long to reach us,” the internal document states. “Filling out this form makes sure that a real human on the team will get to see your application.”
The existence of the form highlights a tension between Google’s public sales pitch for AI-powered hiring products and the caution exercised by its own researchers who understand the technology’s limitations. Google’s Workspace division actively promotes its latest AI capabilities as tools that can “save HR time by quickly creating drafts for job postings, evaluating resumes, and forecasting hiring needs.”
A Google DeepMind spokesperson pushed back against the suggestion that the company’s systems filter out candidates incorrectly. The spokesperson said the organization’s objective is to “recruit and hire the most qualified talent at Google DeepMind” and characterized the form as a way to route applications directly to team members rather than a workaround for flawed technology.
“This team set up a special form to go past the recruiter review, and get their resumes direct to the people on the team,” the spokesperson said. “But there are no shortcuts to getting hired.”
AI-driven hiring tools have spread rapidly across corporate America, but the way companies deploy them varies widely and is often opaque to job seekers. Some employers use AI models to rank applicants, while others rely on automated systems to scan CVs for specific keywords and filter candidates accordingly. The lack of transparency has fueled anxiety among some human resources professionals who worry that automation could eventually reduce the need for certain hiring roles.
These systems have also drawn scrutiny for potential bias. A Bloomberg investigation previously found that OpenAI’s ChatGPT displayed signs of potential discrimination based on applicants’ names. Workday Inc. (WDAY), which sells workplace management software, is facing a lawsuit alleging that its AI hiring systems screen applicants based on race, age, and disability in violation of the law. Workday has denied the allegations, maintaining that humans make final hiring decisions, and did not respond to an additional request for comment.
While some companies use AI to filter candidates, a growing number of job seekers are deploying the same technology to gain an edge. Applicants are using large language models to game keyword-based filters or to churn out applications at record speed. The Google DeepMind team appears to be on guard against this tactic as well. The form designed to bypass automated screening includes an advisory that reads: “A real human will read these. These humans get really tired of reading LLM answers, because they all sound very samey.”
The episode illustrates the broader challenges facing the recruitment technology market, which is projected to grow substantially as employers seek efficiency gains. It also underscores a recurring dilemma for AI developers: the tools they build for customers do not always meet the standards their own engineers demand. For job seekers navigating an increasingly automated hiring landscape, the message from one of the world’s most influential AI labs is clear: when applying to the people who build the algorithms, make sure a human actually sees your resume.