Critics warn AI safety focus overlooks bias and misinformation issues

MIT Technology Review reports that the AI safety conversation is increasingly fixated on rare, catastrophic scenarios while overlooking the everyday harms that already affect millions of users. The article points to a pattern where companies tout modest safeguards—like a chatbot’s promise not to encourage suicide or a social platform’s effort to block children from buying fentanyl—while their broader products continue to generate bias, misinformation, and labor exploitation.

The shift toward high‑profile disaster narratives began as AI tools entered mainstream use in 2022‑2023. Public scares over deep‑fake videos, autonomous‑vehicle accidents, and the potential for AI‑driven weaponry captured headlines and attracted billions in research funding. Governments and philanthropists responded with panels on “existential risk” and large grants for alignment research. In that climate, low‑bar safety claims become easy selling points, even as the same systems amplify existing social problems.

Critics say the focus on rare catastrophes lets companies meet the minimum regulatory checklist while ignoring more common, measurable harms. A recent study found that recommendation algorithms on major platforms increased the spread of health misinformation by 30 percent during the pandemic. Meanwhile, language models trained on biased data continue to produce sexist and racist outputs, affecting hiring and loan decisions. Labor groups note that gig‑economy workers using AI‑assisted tools face lower wages and fewer protections. When firms announce that their chatbots will not suggest self‑harm, they meet a narrow safety metric, but they rarely address how those bots may reinforce harmful stereotypes or erode privacy.

Policy makers are beginning to push back. The European Union’s AI Act now includes provisions for “high‑risk” systems that cover bias and transparency, not just physical safety. In the United States, congressional hearings slated for early 2025 will hear from consumer‑advocacy groups demanding broader accountability standards. Experts argue that a balanced approach—one that treats everyday harms with the same urgency as existential threats—will be essential to protect users as AI becomes ever more embedded in daily life.