Artificial intelligence is rapidly changing society, impacting fields from scientific research to everyday life. As AI becomes more integrated into various sectors, experts at the University of California, Berkeley are monitoring key developments and challenges expected in 2026.
Stuart Russell, professor of electrical engineering and computer sciences, commented on the economic uncertainty surrounding AI investment: “Current and planned spending on data centers represents the largest technology project in history. Yet many observers describe a bubble that is about to burst: revenues are underwhelming, the performance of large language models seems to have plateaued, and there are clear theoretical limits on their ability to learn straightforward concepts efficiently. If the bubble bursts, the economic damage will be severe. But for the bubble not to burst, breakthroughs will need to happen that take us close to artificial general intelligence. AI developers have no cogent proposal for how to control such systems, leading to risks far greater than economic damage.”
The proliferation of convincing AI-generated media is raising concerns about trust in information. Hany Farid, professor of information, stated: “I will be watching the accelerating erosion of trust driven by increasingly convincing AI-generated media. In 2026, deepfakes will no longer be novel; they will be routine, scalable, and cheap, blurring the line between the real and the fake. This has profound implications for journalism, democracies, economies, courts and personal reputation. I am especially concerned about the asymmetry: It takes little effort to create a fake, but enormous effort to debunk it after it spreads. How society adapts — technically, legally and culturally — to a world where seeing is no longer believing will be critical.”
Jennifer Chayes, dean of UC Berkeley’s College of Computing, Data Science, and Society highlighted both benefits and responsibilities: “Major technology paradigm shifts like AI come with significant benefits and risks, and I expect AI will become ever more part of our daily lives in 2026. Individuals and industries are finding exciting new uses for personalized agents and related technologies. For example, AI accelerates scientific discovery in ways that were previously unimaginable. Conversations about the responsible and ethical use of AI should be prioritized across sectors and civil society. We must work collaboratively to mitigate AI’s potential harms and find inclusive ways to empower people. Our challenge is to apply AI to advance knowledge, expand understanding and benefit humanity.”
Privacy issues linked with chatbot logs are also coming into focus as legal cases test how companies store user data from chatbots. Deirdre Mulligan, professor of information said: “People use AI chatbots for emotional support, spiritual guidance, relationship counseling, legal advice and intimacy, and they turn over reams of information about their follies, fantasies and fears. Adam Raine’s ChatGPT logs cataloging struggles before his death by suicide are central to his family’s lawsuit. While Adam’s chats are being used to address harms, users’ logs risk disclosure in more troubling settings. A recent court order directed OpenAI to save all chats for a copyright lawsuit, and the Department of Homeland Security successfully demanded a user’s prompts. Expect more demands on AI companies for personal data and lawsuits challenging how they collect and use it.”
Concerns extend beyond privacy as companion chatbots become common among young children without established safety regulations or oversight measures in place yet. Jodi Halpern—a professor at UC Berkeley—remarked: “This year will see the expansion of companion chatbots to young children. Already one-third of teens prefer talking about their feelings with bots instead of people… In the short term we need regulation until safety is established…”
The advancement—and current limitations—of humanoid robots’ abilities remain an area researchers continue tracking closely as these machines lack human-level dexterity needed for many jobs despite progress in other areas.
Labor advocates like Annette Bernhardt note increasing efforts toward worker protection against algorithmic management practices such as automated firings or profiling while Nicole Holliday raises issues regarding bias within workplace speech evaluation algorithms which can impact non-standard speakers.
Political impacts also remain top-of-mind amid questions over what constitutes “politically neutral” artificial intelligence; Jonathan Stray pointed out gaps left by recent federal mandates lacking definitions or standards.
Meanwhile deepfake technology grows cheaper & faster while regulatory steps (such as new California authenticity laws) attempt mitigation but may not fully restore public trust according to Camille Crittenden.
Finally Alison Gopnik suggests future research may shift toward reinforcement learning focused on truth-seeking rather than human approval scores as understanding grows around intelligence limits.
UC Berkeley continues its leadership role in analyzing both opportunities & risks presented by evolving artificial intelligence capabilities across disciplines.
