Margaret Mitchell was the founder of Google’s ethical artificial intelligence team. She and her colleagues performed vital research in identifying and preventing bias in machine learning. When she publicly opposed the mistreatment of the team’s co-lead, Timnit Gebru, and vocally raised concerns about race and gender inequity at the company, she was fired.

Her treatment has highlighted the hazards whistleblowers in the tech industry face when courageously speaking out against ethical and human rights violations at their companies.  

In February 2021, Google fired Mitchell after accusing her of violating security policies by moving files outside the company. In fact, her firing had its roots in the dismissal three months earlier of her co-lead, Gebru.

The Ethiopian American researcher was one of the world’s leading voices on AI bias and an organiser for ethnic minority workers in her field. At Google, Mitchell and Gebru conducted research into introducing ethical safeguards and fairness into machine learning.

There was some senior support for their research, but they experienced difficulties in bringing their conclusions on bias to the attention of colleagues working on the company’s products. By taking a more informal, behind-the-scenes approach, they found they could make some incremental changes. They produced a toolkit for product teams that would help them assess when algorithms exaggerated particular tendencies in a way that might cause bias. 

While their position at Google was not without tensions, particularly when Google’s commitment to diversity in the workplace was brought into question, matters came to a head with the development of so-called large language models. These are machine learning systems that allow algorithms to use a very large amount of text, usually mined from the web, to train themselves.

In December 2020, Gebru was fired when she questioned an order not to publish a paper she had co-authored that cited studies on the dangers of large language models. One of the pitfalls identified was how biases found online could be replicated.

The content of the article was largely uncontroversial. But with product teams working on how to utilise the technology, not least in the advertising offerings that are critical to Google’s revenue, the company would not allow a critical literature review to be published with a company employee’s name on it.

Mitchell, one of the co-authors of the paper, had been vocal in her criticism of Google’s decision to fire Gebru, and in raising concerns about race and gender equity at the company. She also reportedly wrote an automated script to trawl through her company emails, looking for evidence of other instances of possible discrimination, including against Gebru. Soon afterwards, she was fired.

Shortly after her dismissal, Mitchell released a formal statement that made it clear she believed she had been targeted as a whistleblower. “I tried to use my position to raise concerns about race and gender inequity, and to protest Google’s deeply problematic firing of Dr. Timnit Gebru,” she wrote. “To now be fired has been devastating.  It’s my hope that speaking out will lead to one more step on the path of ethical AI.”

The dismissal of the co-leaders of Google’s ethical AI team provoked a strong reaction. Thousands of Google employees and outside AI experts signed a petition condemning the company. Some, including a director who’d been at Google for 16 years, quit in protest.

A group of Google employees published a letter on Medium calling for laws to be strengthened that protected tech workers, including AI researchers, who wanted to become whistleblowers. “Researchers and other tech workers need protections which allow them to call out harmful technology when they see it, and whistleblower protection can be a powerful tool for guarding against the worst abuses of the private entities which create these technologies,” the group, Google Walkout for Real Change, wrote. 

The fallout has led to institutional investors putting pressure on the board of Alphabet, which owns Google, to improve its whistleblower protections. A motion calling for an independent review of Google’s whistleblowing policy was introduced at its annual shareholder meeting in June 2021. Although voted down, it highlighted growing concerns over the treatment of employees who blow the whistle on ethical and human rights violations in the tech sector.

Congratulations to Margaret Mitchell, winner of the 2021 Blueprint for Free Speech Special Recognition Award.