Smart Glasses Raise Privacy Fears as AI Spending Surges
Smart glasses are intensifying privacy risks in India while AI companies face pressure to justify nearly $1.1 trillion in infrastructure spending.
Smart glasses, massive data-center investments and increasingly unusual AI experiments are pushing familiar technology debates into new territory. In India, wearable cameras are creating immediate risks for people recorded without consent. At the same time, the companies building AI infrastructure face difficult questions about whether future earnings can justify their spending.
Other developments—from alleged cyberattacks and lower-cost AI models to pain-detection apps—show how quickly the technology landscape is expanding, often faster than social norms and safeguards can adapt.
Smart glasses create new privacy risks in India
The privacy implications of smart glasses are becoming tangible in India. Shubnam, who attended a protest in Delhi, later discovered that a content creator wearing Meta smart glasses had recorded them without their knowledge. The resulting Instagram reel attracted millions of views, along with transphobic abuse and AI-generated memes.
The episode illustrates how wearable cameras can make recording less obvious than filming with a phone. As smart glasses become more common, experts expect more people to face similar experiences.
India presents particularly acute concerns because covert recording and the nonconsensual circulation of images are already widespread. Smart glasses could make those practices easier by placing cameras inside products that resemble ordinary eyewear.
The risks extend beyond viral videos and online pranks. Smart glasses are also emerging as a police surveillance tool, raising broader questions about when people are being recorded, how footage is used and whether meaningful consent is possible in public spaces.
AI infrastructure faces a trillion-dollar test
The economic case for the AI boom depends partly on whether a small group of major technology companies can generate enough revenue from their infrastructure investments.
Jessica Wachter, a finance professor at the University of Pennsylvania, approached the issue by examining the spending of so-called hyperscalers. Rather than attempting to forecast how widely AI models will be adopted, she considered how quickly these companies’ earnings would need to rise to justify their planned data-center spending.
Expenditures through 2027 are expected to reach nearly $1.1 trillion. According to the analysis highlighted by MIT Technology Review, AI companies would need an extraordinary productivity increase simply to break even by 2030.
That framing shifts attention from the capabilities of individual models to the financial assumptions supporting the industry. Building data centers at this scale requires confidence that demand, revenue and productivity gains will grow enough to support the investment.
Security, competition and policy reshape the industry
Several developments show that AI’s expansion is unfolding alongside cybersecurity threats, pricing pressure and political scrutiny:
- The ShinyHunters hacking group claimed it stole more than 2 terabytes of data covering almost all FBI employees. The group said the material included agents’ names, addresses and phone numbers, and described the alleged attack as retaliation for an FBI alert.
- Anthropic and OpenAI released lower-cost models as they face competition from cheaper Chinese alternatives. Startups have also been reducing their reliance on the two AI labs by adopting open models.
- The CEOs of Anthropic and OpenAI were scheduled to brief the UN Security Council amid continuing debate over AI safety.
- US Treasury Secretary Scott Bessent was reportedly being considered for an AI policy role after playing a central part in US-China AI discussions. Michael Kratsios and Scott Kupor were also identified as contenders.
- President Donald Trump told the UN General Assembly that he wanted AI to be called “superintelligence,” arguing that the word “artificial” makes the technology sound fake.
Together, these developments underline how AI competition now spans model prices, national policy, international security and control over sensitive data.
New systems combine biology, light and automation
Research and commercial projects are testing unconventional ways to improve computing, transport and food production.
Data centers are beginning to replace copper connections with photonics, which can transfer information using light while producing less heat than electrical wiring. Virtual power plants have also been proposed as one way to help meet data centers’ electricity needs.
AWS, meanwhile, is using rat neurons in work intended to make video AI faster. The project moves the idea of computing systems built partly from biological material closer to practical experimentation.
Beyond computing, scientists are working on methods that use renewable electricity to produce food while requiring less land. Companies are also developing processes described as creating food from air.
Transportation companies are pursuing hybrid approaches rather than relying entirely on automation. Uber is betting that human drivers can complement robotaxis by covering periods of high demand while autonomous vehicles recharge.
AI tools seek to measure pain
AI is also changing attempts to assess pain in people who cannot communicate verbally. At Orchard Care Homes, nurses previously used an observational scale for residents in this situation. Agitation was sometimes treated as a behavioral problem even when untreated pain was the cause.
In 2021, the care-home chain began testing PainChek, a smartphone application that scans faces for microscopic muscle movements and produces a pain score using AI. Within weeks, the pilot unit reported fewer prescriptions and calmer corridors.
Researchers are now trying to make pain measurable by cameras or sensors with reliability comparable to a blood-pressure reading. The work could provide clinicians with another source of information, but it also raises questions about how algorithmic scoring may affect the way suffering is understood and treated.
A widening gap between capability and safeguards
The latest developments show technology advancing across surveillance, infrastructure, healthcare and biological computing. The central challenge is no longer only whether these systems work. It is also whether privacy protections, financial returns, safety practices and public oversight can keep pace with their deployment.
Attribution: revew
Originally reported by revew.