The tech industry is finally listening to calls for more inclusive AI—but at what cost? As Howard University and Google launch Project Elevate Black Voices, the spotlight is now on collecting African American English (AAE) data to improve AI systems. While this is a much-needed move to correct decades of bias in tech tools, it also raises a crucial question: Who’s protecting Black voices in AI as they become a powerful asset for Big Tech?
The Promise: Elevating Black Voices with Purpose
The partnership between Howard University and Google is collecting over 600 hours of speech data across 32 states to represent the wide spectrum of African American English. This data can improve tools like Zoom captions, AI notetakers, and transcription software that often fail to understand AAE. From healthcare to legal sectors, these upgrades could make workplace tech more accurate and inclusive—especially for Black professionals who’ve had to “code-switch” just to be understood.
The Peril: Historical Exploitation Can’t Be Ignored
Despite its promise, many remain skeptical—and for good reason. The tech industry has a long record of extracting from Black communities without credit or compensation. Black creators have sparked viral trends only to be erased, while algorithms disproportionately flag their content or misidentify them in surveillance systems. Critics fear this dataset could be misused for digital Blackface, commercial AI avatars, or surveillance tools that further harm the very people it's meant to help.
The Tension: Empowerment or Exploitation?
Howard has committed to licensing the dataset exclusively to HBCUs—for now. But questions remain: Who will access this data next? Will the community benefit financially or culturally? Even within Black-led spaces, harm can still happen. As Audre Lorde warned, oppression can be replicated from within. To truly protect Black voices in AI, we must build systems rooted in community power—not just corporate interest—and ensure that AI tools are built with Black communities, not just about them.
The Path Forward: Equity Requires Vigilance
There’s no doubt that the Howard African American English Dataset can help make AI better, fairer, and more inclusive. But we must remain cautious. Protecting Black voices in AI means setting boundaries, demanding transparency, and ensuring that Black researchers, workers, and communities are not only represented—but respected, compensated, and in control. Tech may be evolving, but without accountability, history will repeat itself—this time coded into the machines we use every day.
Comment