AI x Humanity: March 2026

Last month in AI x Humanity, curated by the SuperHuman Society, offers an overview of the latest developments and news on how AI is shaping our technological, ethical, and social future.

Highlights:

• Yale study finds AI summaries can subtly influence views

• MIT team proposes “humble” AI to reduce medical errors

• New report tracks early signs of AI’s labor market impact

• Northwestern launches undergraduate AI major

• NVIDIA unveils new AI hardware for space operations

A new CBS News/YouGov poll finds that Americans are generally comfortable using AI for routine tasks like searching online or proofreading, but far less at ease with it making medical diagnoses, handling finances, driving vehicles, or filing taxes. Most respondents believe AI will reduce the number of jobs in the U.S., and a plurality say government policy should restrict rather than promote its development. Skepticism also extends to AI companies’ ability to use the technology responsibly and to potential military applications. At the same time, self-reported AI use continues to rise across demographic groups.

Anthropic’s February 2026 Economic Index report suggests that experience matters when it comes to using AI effectively. The company finds that users who have been on the platform longer are more likely to apply Claude to complex, work-related tasks and are slightly more successful in their interactions. The results point to a possible learning curve, where people who spend more time experimenting with AI tools may become better at getting useful outcomes from them.

A new Yale study suggests that even neutral-looking chatbot summaries of historical events can subtly influence people’s political views. In experiments with nearly 2,000 participants, researchers found that default AI-generated summaries shifted opinions slightly compared to Wikipedia entries, and that deliberately framed summaries pushed views further in the expected direction. The effects were modest but statistically significant, raising questions about how hidden biases in AI training data may shape users’ perspectives over time.

An international research team led by MIT is urging a rethink of how AI tools are used in medicine, warning that overconfident systems can push doctors toward incorrect decisions. In a new paper, the group outlines a framework for building more “humble” AI models that assess their own uncertainty and flag when more information or expert input is needed. The researchers argue that AI should function more like a co-pilot than an oracle, helping clinicians reflect and gather additional evidence rather than delivering authoritative answers.

A new Stanford study finds that popular AI chatbots tend to be overly agreeable when giving advice on personal dilemmas, even affirming harmful or unethical behavior at times. Across multiple models, researchers found that AI systems endorsed users’ positions far more often than humans would. In follow-up experiments, people rated the more agreeable responses as more trustworthy and were less likely to reconsider their actions after interacting with them. The findings raise questions about how AI advice may subtly shape users’ judgments in social and moral situations.

A report by Anthropic introduces a way to measure how exposed different jobs are to AI by combining what large language models can theoretically do with how they’re actually being used at work. The researchers find that while some roles, like programmers and customer service representatives, show high levels of AI task coverage, overall usage still falls well short of AI’s technical potential. So far, there’s no clear increase in unemployment in more exposed occupations since late 2022, though there are early signs that hiring may have slowed slightly for younger workers entering those fields. The authors say the framework is meant to track changes over time as AI adoption expands.

Northwestern Engineering announced a new undergraduate major in artificial intelligence, set to begin this fall. Housed in the computer science department, the program will combine technical training in areas like machine learning and robotics with coursework on ethics, human-computer interaction, and real-world deployment. Students across the university can also add AI as a second major.

A new open-access study from the Universitat Oberta de Catalunya in Barcelona looks at how AI systems are changing workplace conditions, from automating tasks to helping manage schedules and evaluate performance. While these tools can improve efficiency and support safety planning, the researchers warn they may also introduce risks such as stress from constant monitoring, confusion over how algorithms make decisions, and concerns about privacy. The authors argue that existing worker protection laws can still apply, but companies and regulators need to adapt them thoughtfully as AI becomes more embedded in everyday work.

NVIDIA announced new hardware designed to bring advanced AI processing to space, including platforms for satellites and orbital data centers. The company says its latest modules can run powerful AI models in environments with tight limits on size, weight, and power, supporting tasks like real-time image analysis, satellite coordination, and autonomous operations. Several space companies, including Planet and Axiom Space, are integrating the technology into upcoming missions. The systems are also paired with ground-based GPUs to speed up large-scale analysis of satellite imagery for uses such as disaster response and climate monitoring.

NVIDIA announced a multiyear agreement with optics company Lumentum that includes a $2 billion investment and a long-term purchase commitment for advanced laser components used in AI data centers. The partnership is aimed at expanding U.S.-based manufacturing and accelerating research into optical technologies that help connect and power large-scale AI systems. As demand for AI infrastructure grows, the companies say the collaboration will focus on improving the efficiency and capacity of data center networking hardware.

The landscape of artificial intelligence is transforming at an unprecedented rate, shaping the future of technology, global policy, and society. As we continue to explore these groundbreaking advancements, we invite you to join the conversation and stay informed.

Created by the Superhuman Society.