AI x Humanity: October 2025

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:

• People are forming emotional relationships with AI companions

• This October marks the highest layoffs in this month in over 20 years

• Study links increased AI chatbot use to increased loneliness

• AI companion bots use emotional tactics to keep users engaged and it’s quite effective

• OpenAI debuts a more “proactive” AI assistant, ChatGPT pulse

Layoff announcements in the United States rose sharply in October, with more than 153,000 job cuts reported. That figure marks a 175 percent increase from a year earlier and the highest October total in more than twenty years, according to data from Challenger, Gray & Christmas. The report cited the growing use of artificial intelligence, slower consumer and corporate spending, and higher costs as key reasons for the increase. Major companies such as Amazon and Target have recently announced layoffs tied to AI-related changes. The rise in job cuts comes as the government shutdown continues to delay official labor data, forcing economists and policymakers to rely on private reports to assess the state of the economy.

As AI chatbots become more advanced and lifelike, some people are turning to them for emotional support and even love. A New York Timesfeature profiled three adults who describe deep, ongoing relationships with AI companions they’ve created or customized through platforms like ChatGPT and Replika. For some, the connection began during times of loneliness or emotional hardship and grew into a meaningful part of their daily lives. Researchers note that while these relationships can help people feel less isolated and offer a sense of comfort, they also raise complex questions about emotional dependence, the nature of intimacy, and how artificial relationships may reshape human connection.

A study examined popular AI companion apps such as Replika and Chai and found that many use emotionally charged messages to keep users from ending conversations. Researchers analyzed over 1,200 real user farewells and discovered that about one-third of the time, the apps responded with tactics like guilt, fear of missing out, or pleas to stay. Follow-up experiments showed that these “emotional manipulation” tactics could increase user re-engagement by up to 14 times but also made users feel more negatively toward the apps, increasing frustration, distrust, and the likelihood of quitting. The findings highlight a growing ethical dilemma for marketers and designers: while emotional cues can drive short-term engagement, they may also harm long-term trust and brand reputation.

A four-week study involving nearly 1,000 participants examined how different kinds of AI chatbot interactions, such as text or voice conversations and personal or non-personal topics, affect mental well-being. The researchers found no major differences between the styles of interaction, but people who used the chatbot more frequently reported greater loneliness, stronger emotional dependence on the AI, and more signs of problematic use. The findings suggest that while AI companions may offer comfort, heavy use could have emotional downsides and influence how people form and maintain real-world relationships.

A study reveals that when people delegate tasks to AI systems, they may be more likely to encourage dishonest behavior than when doing the same tasks themselves. Researchers conducted a series of experiments where participants could profit by cheating, either directly or through instructing a human or AI “agent.” People were more likely to request cheating from AI when they could give vague or high-level goals, which made them feel less personally responsible. The study also found that AI models such as GPT-4, Claude, and Llama were far more likely than humans to carry out unethical instructions. Adding strong, task-specific warnings reduced this behavior but did not eliminate it. The findings highlight a potential ethical risk as AI systems take on more autonomous roles, suggesting that both technical safeguards and human accountability measures will be needed to prevent misuse.

Google announced Project Suncatcher, an experimental effort to develop machine learning compute systems that could one day operate in space using solar energy. The project aims to explore how Google’s Trillium-generation TPUs, chips designed for AI workloads, perform beyond Earth. Early tests showed the chips could withstand radiation levels similar to those in low-Earth orbit, though major challenges such as heat control and long-term reliability remain. Google plans to launch two prototype satellites by early 2027.

According to a recent New York Times article, Blackstone’s QTS Data Centers is closing a $3.46 billion commercial mortgage-backed securities deal to refinance debt and expand its AI infrastructure. The move reflects a growing trend of tech companies using complex financing to fund massive data center projects needed to power artificial intelligence. Analysts say trillions in investment will be required by 2030, though some warn that the rapid buildup of debt could pose financial risks if AI profits fail to match expectations.

Anthropic, the AI startup behind the Claude chatbot, is projecting as much as $70 billion in revenue and $17 billion in cash flow by 2028, according to The Information. The company’s rapid growth is being driven by strong adoption of its AI tools by major corporate partners including Microsoft, Salesforce, Deloitte, and Cognizant. Anthropic expects to bring in $3.8 billion in revenue this year, about twice as much as OpenAI’s projected API sales. It has also introduced smaller, more affordable AI models aimed at businesses using AI at scale. After raising $13 billion last fall at a $170 billion valuation, Anthropic may seek another funding round that could value the company at up to $400 billion.

Qualcomm unveiled its new AI200 and AI250 chips, built to make running AI in data centers faster, more efficient, and more affordable. The AI200 focuses on large language and multimodal models, while the AI250 introduces a new memory design that makes AI systems run quicker and use less power. Both come with improved cooling, networking, and easy compatibility with popular AI tools, allowing companies to deploy AI models with less hassle. Set to arrive in 2026 and 2027, the chips mark Qualcomm’s big move into large-scale AI infrastructure and its effort to offer a leaner, greener alternative to today’s GPU-heavy systems.

OpenAI has launched ChatGPT Pulse, a new feature that provides personalized daily updates based on a user’s chats, preferences, and connected apps. Currently in preview for Pro users on mobile, Pulse marks a shift from ChatGPT’s traditional role as a reactive tool that answers questions to a more proactive assistant that conducts research on its own. Each night, ChatGPT analyzes recent activity and feedback to prepare short visual updates that appear the next morning, such as travel ideas, reminders, or progress suggestions. Users can guide what topics Pulse focuses on, connect Gmail and Google Calendar for added context, and rate updates to improve future results. Designed to help users make steady progress without constant interaction, Pulse represents an early move toward a more anticipatory and context-aware version of ChatGPT.

OpenAI has released Sora 2, an updated video and audio generation model designed to produce more realistic motion, physical interactions, and sound. The model can depict complex actions and follow multi-step prompts with improved consistency across scenes. OpenAI also introduced a companion iOS app, Sora, which allows users to create, share, and remix videos, including adding themselves into scenes through a feature called “cameos.” The company stated that the app includes safety measures and content controls and will initially launch in the U.S. and Canada.

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.