Bias & Fairness in AI Models
Research Rundown #156, plus new memos on Loft Orbital, Base Power and more.
Research Rundown
What makes AI models fair? What makes them unbiased? Are fairness and bias characteristics that describe the data used to train models, the way models are trained, or only the outputs of models? As both personal and enterprise AI model usage have increased, questions about the fairness and intrinsic bias of AI model outputs have grown faster than answers. Users have flagged instances of AI racism in image generation, ageism in job application reviews, and more nebulous issues with AI values in models that consistently validate users, even when they are factually or morally in the wrong.
While these questions are central to debates about artificial intelligence, competing metrics demonstrate that fairness in AI is not a single, universally agreed-upon concept. Rather, its definition relies on differing intuitions about equity, while many popular definitions of fairness in AI discourse are mutually incompatible.
Read our full deep dive on how AI model creators and users are thinking about evolving and evaluating model bias and fairness here.
Loft Orbital sits at the intersection of aerospace and software, abstracting the hardest parts of getting to orbit. Check out the full memo here to learn more.
Base Power’s tech could be the secret behind the next wave of clean energy. Find out more in our memo here.
At the heart of digital creativity, Figma is redefining how teams bring ideas to life. Explore our memo here for all the details.
Anysphere, the company behind code-assistance tool Cursor, is considering investment offers valuing the company at $30 billion. Anysphere closed its Series C in June 2025 at a valuation of $9.9 billion, and reportedly rejected investment offers over the summer that would have valued that company at $18-22 billion. Secondary shares valuing the company around $30 billion have already occurred, according to The Information.
Microsoft and Harvard Medical School announced an update to the Copilot chatbot that will integrate Harvard Health Publishing data into Copilot responses to health-related queries. This partnership is considered a bid by Microsoft to diversify away from OpenAI. Microsoft noted it is also working on Copilot functionality to find healthcare providers based on insurance and location.
Sam Altman shared a preview of OpenAI’s vision for the future of ChatGPT as an operating system at OpenAI’s Developer Day on Monday of SF Tech Week. Available now in beta for some developers, ChatGPT now features chat-interface apps, including Spotify, Canva, and Zillow, with plans to allow “agentic commerce protocol” for developers to monetize their apps.
Oracle stock fell as much as 7% on Tuesday after a report published by The Information revealed that Oracle’s AI Cloud Compute for-rent generated only a 14% margin in the third quarter of 2025, lower than analyst expectations. The stock ended the day down only 3% and has since fully covered the Tuesday losses.
Jeff Bezos, Jamie Dimon, the Bank of England, and the IMF all raised concerns this week about the possibility of an AI bubble. Bezos called AI a “kind of industrial bubble” at a tech conference in Italy, saying “The [bubbles] that are industrial are not nearly as bad, it can even be good, because when the dust settles and you see who are the winners, societies benefit from those inventions.”
AMD and OpenAI announced a partnership aiming to deploy six gigawatts of AMD Instinct GPUs, the company’s product line designed for deep learning, neural networks, and general-purpose GPU performance. This partnership will begin with one gigawatt of AMD Instinct MI450 GPUs in the second half of 2026. The deal includes a warrant for OpenAI to purchase up to 160 million AMD shares, equivalent to about 10% ownership in the company.
Tesla released news on Tuesday that it plans to offer less expensive versions of its Model Y and Model 3, priced at $40K and $37K, respectively. EV analysts have pointed to this announcement as a response to Congress’s elimination of a $7.5K tax credit for electric vehicles last month. This announcement follows Tesla’s plans to release a $25,000 vehicle manufactured in Mexico last year, a target that was eventually nixed to focus on autonomous vehicles and robotics.
China has announced expanded restrictions on exports of rare earth elements required for semiconductor manufacturing, adding five new elements to the restricted export list and launching an investigation into Qualcomm. The investigation’s focus is a potential antimonopoly law violation tied to Qualcomm’s acquisition of Autotalks, a maker of semiconductors designed for car crash prevention systems. China similarly accused Nvidia of violating antimonopoly law last month, related to its acquisition of Mellanox Technologies.
At Contrary Research, we’ve built the best starting place to understand private tech companies. If you're interested in researching and writing about tech, apply here for our Research Fellowship!