Amazon, Alphabet, Microsoft, Meta probed by lawmakers on use of AI ‘Ghost’ staff

Representational Image for Artificial Intelligence; Man typing on keyboard with a green background (Image: Reuters)

Democratic lawmakers are pressing the top tech firms to open up about the conditions of their “ghost work” — unseen laborers like those labeling data and rating responses who have become pivotal to the artificial intelligence boom.

“Despite the essential nature of this work, millions of data workers around the world perform these stressful tasks under constant surveillance, with low wages and no benefits,” a group of lawmakers led by Massachusetts Senator Ed Markey and Washington Representative Pramila Jayapal wrote Wednesday in a letter to the chief executives of nine companies, including Amazon.com Inc., Alphabet Inc., Meta Platforms Inc., Microsoft Corp, and International Business Machines Corp.

“Workers are expected to screen out dangerous chatbot answers, but they may have little time to assess an answer’s safety,” they added. “Data workers are often given scant training or supervision, which can result in the introduction of bias.”

The lawmakers ask the executives to answer extensive questions about their data workforces, including about their ability to take breaks, appeal suspensions or access mental health resources when facing traumatizing content. “Tech companies must not build AI on the backs of exploited workers,” write the Democrats, including Massachusetts Senator and former presidential candidate Elizabeth Warren.

Along with the established tech giants, the letter’s recipients include the newer AI-focused firms OpenAI Inc., Inflection AI, Scale AI Inc. and Anthropic.

To develop AI products, US companies rely extensively on sub-contracted staff, located domestically or abroad, who are hired via outside staffing firms and often lack the benefits provided to firms’ own direct employees. Companies also rely on such work for other burdensome tasks, such as content moderation and product quality assurance.

For generative AI tools, which produce answers to text prompts in the form of text, photos or even videos, the thousands of unseen contract workers are generally hired to train, fix and improve algorithms, which are then presented to customers as operating via the magic of technology. They are frequently underpaid and say they are stressed and overworked.

In some cases, workers report trauma from viewing disturbing images in order to filter them out. OpenAI paid workers in Kenya less than $2 an hour to keep that kind of content out of ChatGPT, Time reported in January. Google’s ChatGPT rival, Bard, also relies on the labor of humans who told Bloomberg they are given little training, high workloads and poor pay.

The lawmakers’ letter comes as executives including the leaders of Tesla Inc., Meta, Microsoft and Alphabet are slated to meet senators Wednesday afternoon at a closed-door AI summit hosted by Senate Majority Leader Chuck Schumer, who was not among those signing the document.

“These tech moguls are under-paying workers, failing to provide them basic protections and benefits, and subjecting them to an extensive web of surveillance in order to prop up their business,” Markey said in an emailed statement. “When they come to the Capitol to tout their innovation and excellence, I’d like to hear them answer for these disgusting labor practices.”