July 10, 2026 — by Lulu Liu in Uncategorized
My blogs have been exploring the problem of algorithmic bias, the structural problem ingrained in our systems that quietly shapes our daily lives. Most of what I have pointed out so far has been academic research and investigative journalism. This time, I will be focusing on something different: the AI Now Institute, a nonprofit research organization at the intersection of policy, advocacy, and public education. And we’ll be looking closely at their most recent annual report, Artificial Power: 2025 Landscape Report.
Image via AI Now
The AI Now Institute was founded in 2017 by researchers Kate Crawford and Meredith Whittaker, and you can read more about their funding model and mission on their About Us page. An interesting detail about this organization is that as of 2022, it explicitly refuses funding from corporate donors, including the tech companies its research scrutinizes. That is a move worth noting; their credibility rests on both the quality of their research and also on the transparency of who is not paying for it. For a public audience trying to figure out who to trust regarding AI issues, this distinction matters.
As a website, AI Now operates in the policy research genre: a hybrid space between academic publishing and advocacy for the public. Their homepage opens with punchy, declarative statements designed to hook a non-specialist reader. Their framing of the 2025 report is a good representation of this: “Today’s AI isn’t just being used by us — it’s being used by us.” This public argument certainly works. The site also organizes content around clear research areas, including accountability, privacy and surveillance, labor markets. inequality, and more. This gives readers multiple entry points depending on what they already care about.
The Artificial Power report argues that the AI boom has not democratized technology; instead, it has concentrated power further into the hands of a smaller number of large companies such as Microsoft, Google, Meta, and OpenAI. The report makes the claim that AI is predominantly deployed as an infrastructure that shapes access to housing, credit, healthcare, and emplyment, often without the people affected knowing it is happening. If you have read my previous posts on COMPAS and the fairness impossibility theorem, this should feel familiar. The problems I have been writing about at the technical level are symptoms that the larger AI research group is describing.
Image via ProMarket
Where the report is most effective is in how it goes from diagnosis to action. Rather than ending with vague “calls for action” to its audience, Artiticial Power offers detailed recommendations aimed at community organizers, policy makers, and everyday users. They thread the needle between too technical and too vague in such a way that post public-facing AI writing fails to do. That said, the jump from the homepage’s punchy line to the actual content is steep for those without any policy or technical background. More intermediate content like explainer posts, short summaries, and annotated works would go a long way in closing that gap.
For the readers interested in exploring AI issues, even without much knowledge or background, I would strongly recommend bookmarking AI Now and starting with their executive summary of Artificial Power. It connects the technical failures I have been writing about to the broader power structures that allow them to persist and scale. Understanding one without the other only gets you halfway there.
Image via GeeksforGeeks
Leave a Reply