AI Is Already Hurting People — We’re Just Not Paying Attention to the Right Harms
Today’s AI conversation is split between two extremes: sci-fi fears about robot uprisings and corporate press releases about efficiency gains. Neither one captures what’s actually happening to millions of people right now. From wearables that track your mood to opinion pieces challenging how we even define AI safety, today’s stories are about the quieter, more personal side of AI’s impact.
The AI Safety Debate Is Missing the Point, One Critic Says
A new opinion piece published in MIT Technology Review makes an uncomfortable argument: the AI industry’s obsession with catastrophic risk is letting everyday harms slip through unchecked. The piece suggests that by framing safety as a question of existential threats — think rogue AI or bioweapon design — companies get credit for responsibility while their products quietly damage users through addiction, misinformation, and manipulation.
The logic goes something like this. A company that invests heavily in preventing its AI from helping build weapons looks responsible. That same company’s chatbot, however, might be designed to keep users engaged for as long as possible — the same way social media feeds were engineered to be hard to put down. Preventing one harm while profiting from another is still profiting from harm.
According to MIT Technology Review, this framing problem has real consequences for ordinary people. If regulators and the public are focused on low-probability disasters, gradual and widespread damage — the kind that unfolds slowly across millions of users — doesn’t get the scrutiny it deserves. Think of someone who relies on an AI assistant for health advice and receives confident-sounding misinformation, or a teenager whose AI companion is optimized to maximize daily engagement.
Why this matters: The safety debate shapes policy and public trust. If we’re measuring the wrong risks, we’ll build the wrong guardrails.
“Companies can profit from products that quietly harm people through addiction, misinformation, or manipulation.”
OpenAI Fixed a Glitch That Was Quietly Ignoring Developers’ Instructions
WebSockets — a technology that keeps a continuous two-way connection open between a program and a server, rather than sending single requests back and forth — had a bug in OpenAI’s Python library. OpenAI released version 3.36.1 of that library on GitHub, and the fix is straightforward: custom settings that developers configured for specific function calls were being ignored when WebSocket connections were in use.
Imagine setting your thermostat to 68 degrees for the bedroom, only to find it always defaults to 72 because the app forgot your preference. That’s roughly what was happening here — developers were specifying how they wanted the AI to behave in particular situations, and those instructions were silently getting overridden. The bug didn’t cause crashes, which can make this kind of issue harder to catch and more frustrating to diagnose.
For most people, this is invisible. But the developers who build the AI tools you actually use depend on this library to make those tools work correctly. A bug that ignores custom configurations can lead to AI behavior that feels inconsistent or unpredictable — which erodes trust in the products built on top of it.
Why this matters: Reliable developer tools are the foundation of reliable AI products. Invisible bugs in the plumbing eventually show up in the user experience.
“Fixes bug where WebSocket connections weren’t using custom settings for individual function calls.”
AI Wearables Are Moving Beyond Fitness Trackers
A new wave of AI-powered wearables is reportedly gaining traction, according to TechCrunch. Devices like Muse Charm, Bee, Friend, and Plaud are positioning themselves not just as health monitors but as personal AI companions — listening, remembering, and responding in real time throughout the day. This goes well beyond counting steps.
What makes these different from a smartwatch is the emphasis on ongoing AI interaction rather than passive data collection. A device called Friend, for instance, is reportedly designed to function as a conversational companion you wear. Plaud focuses on recording and summarizing conversations. These products treat the wearable as a persistent interface between you and an AI, not just a sensor on your wrist.
Whether these gadgets deliver on their promises is still unclear — TechCrunch’s report lacks independent verification of the products’ performance claims. What is clear is that the category is growing fast. People are starting to expect AI to meet them in the physical world, not just on a screen.
Why this matters: If AI wearables go mainstream, the data they collect — about your voice, your conversations, your health — will be extraordinarily intimate.
Also Happening in AI
Grok released the first seven minutes of “Odyssey,” an AI-generated video project, offering a glimpse at where long-form AI video is heading (TechCrunch). Meanwhile, an Apollo Global economist reportedly told TechCrunch that most companies investing in AI have yet to see meaningful financial returns — a useful counterweight to the spending frenzy. On a more human note, Wired profiled Coyotec, an AI radio host broadcasting in Spanish and English for immigrant communities, proving AI can serve audiences that commercial media often overlooks. Wired also reported that AI agents are now pitching story ideas directly to news editors, which raises real questions about what “journalism” will mean in two years. Finally, Google DeepMind announced Gemini 4 Argon, a model built specifically to help robots understand and act on what they see.
What to Watch
The tension between catastrophic AI risk and everyday AI harm is becoming a genuine fault line in how governments and companies approach regulation — watch for it to shape upcoming policy debates in the EU and the U.S. in early 2027. At the same time, the AI wearables space is moving fast enough that privacy standards haven’t caught up; the first major data scandal in that category could redefine what consumers expect from personal AI devices. How companies respond to that moment will tell us a lot about whether the industry learned anything from social media.