Join on WhatsApp
Join on Telegram
Cairo – During AI Everything MEA Egypt 2026 in Cairo today, Margaret Mitchell of AI firm Hugging Face stepped off the stage from the panel titled “Human-Centric AI: Power, Responsibility, and the Global South” and continued a conversation that felt more urgent than theoretical.
Hugging Face is a US-based open-source AI company that builds and hosts machine learning models, making them accessible to developers, researchers, and startups around the world. It has become a central hub for sharing and testing AI systems.
The panel had examined what responsibility AI builders carry as systems advance faster than regulation, whether open and closed AI models can coexist, and how emerging economies can avoid becoming digitally dependent.
After the panel, when asked whether large AI companies truly take privacy seriously, she didn’t give a blanket answer. “It depends on the company,” she said to the press.
She pointed to Microsoft and Google as places where, from her experience, privacy was taken seriously, while noting that lawsuits and fines can sometimes become a calculated business cost. “You have to really robustly handle privacy,” she said, adding that long-term trust is what sustains companies, not short-term shortcuts.
Mitchell, whose work centers on fairness and algorithmic bias, stressed that ethics often has to move before regulation. “Regulation is tending to lag AI development,” she said.
Read also: Ai Everything MEA Egypt 2026 Kicks Off in Cairo
That gap, she argued, is where AI ethicists help companies and governments weigh trade-offs before harm occurs, not after public backlash.
Bias, she said plainly, harms those already marginalized. AI systems reflect the data they are trained on, and that data disproportionately represents certain demographics.
As a result, healthcare systems, language models, and decision tools can fail more often for women, Black communities, and others who are underrepresented. “A lot of the people who are marginalized end up having more unfair representations,” she said, adding that the problem is structural, not incidental.
On open source AI, a recurring theme of the panel, Mitchell acknowledged the tension. Open systems promote transparency and accountability, but they can also be misused. “There is no such thing as a technology that’s only good without other kinds of bad,” she said.
The task, in her view, is to examine long-term impact and align development with core human values. She noted that openness can act as a check on exaggerated claims. Transparency, she said, helps distinguish technical reality from marketing.
Perhaps her strongest position came on privacy. If developers want to protect users, she argued, encryption should be the standard — and not partial encryption. “If you truly want to preserve privacy, then you have to have encryption that not even the company has access to,” she said.
Mitchell also warned of growing psychological and social risks. Dependency on AI systems, she said, can erode human judgment and critical skills.
As people begin forming emotional attachments to chatbots or relying heavily on automated tools, society risks weakening its own competencies. “It’s important to always have experts who are not working with the assistance of AI systems,” she said, cautioning against what she described as critical skill degradation.
True to the theme of the panel she had just left, Mitchell emphasized that power without responsibility is unstable, and trust — once lost — is difficult to rebuild.
