Google team warns applicants its AI hiring filters may be unreliable

Google's artificial intelligence researchers are warning job applicants that the company's internal hiring systems may incorrectly screen out CVs.

The warning comes as the company promotes AI tools to businesses to streamline recruitment, Bloomberg reports.

Google DeepMind's AGI Safety and Alignment Team, which works on mitigating risks from advanced AI, is asking applicants for some of its open positions to complete a special form alongside their applications.

The form is intended to ensure their CVs are seen by a human member of the team, according to a document reviewed by Bloomberg.

The document, marked “PLEASE DO NOT SHARE THIS DOC WIDELY”, says there is a “non-trivial probability” that a CV could be screened out incorrectly or take too long to reach the team.

Google DeepMind said it aims to recruit the most qualified candidates and denied that its systems incorrectly filter out applicants.

“This team set up a special form to go past the recruiter review, and get their resumes direct to the people on the team. But there are no shortcuts to getting hired,” a spokesperson said.

AI has become increasingly common in recruitment, with some companies using models to rank applicants or screen CVs for keywords. But the use of such systems has raised concerns about transparency and potential discrimination.

Google's Workspace team markets AI features to corporate customers as tools that can help HR departments save time by creating job-posting drafts, evaluating CVs and forecasting hiring needs.

AI hiring systems have also faced scrutiny over potential bias. A Bloomberg investigation found signs of potential bias in OpenAI's ChatGPT based on applicants' names. Workday, which sells workplace management software, is facing a lawsuit alleging that its AI hiring systems screen applicants based on race, age and disability. Workday has denied the allegations, saying humans remain responsible for hiring decisions.

Meanwhile, job seekers are increasingly using AI to tailor applications or produce large numbers of CVs. The Google DeepMind team is also warning applicants against relying on AI-generated responses, saying human reviewers can become tired of reading answers produced by large language models because they tend to sound similar.

The special application form therefore serves both to bypass automated screening and to put applicants directly in front of human reviewers.