The Download: India’s troubling smart glasses and AI’s $1 trillion wager

Welcome to The Download, our weekday newsletter delivering a daily roundup of the most important developments in technology.

Smart glasses are already causing havoc in India

Shubnam discovered they had been secretly filmed at a protest in Delhi only after spotting themselves in an Instagram video—captured by a content creator wearing Meta smart glasses. The mocking reel racked up millions of views and brought a wave of transphobic abuse alongside AI-generated memes.

Specialists caution that Shubnam's experience will become increasingly common as smart glasses reach mass adoption. The danger is especially pronounced in India, where non-consensual filming and the spread of images have long been widespread problems.

And the concern goes beyond ordinary people being turned into unwilling stars of viral "pranks." Law enforcement is also deploying these devices as a surveillance instrument.

Read more about why smart glasses are creating new privacy risks in India.

—Anuj Behal

MIT Technology Review Narrated: AI's trillion-dollar gamble

Jessica Wachter, a finance professor at the University of Pennsylvania, set out to measure what AI might mean for the economy in the coming years. Her starting point was straightforward: a small group of hyperscalers is pouring enormous sums into constructing AI data centers.

Rather than guessing how broadly AI models will actually be adopted, Wachter calculated how quickly these companies' revenues would have to climb to make their investments worthwhile by 2027, when spending is projected to hit nearly $1.1 trillion.

The findings are striking: for AI firms merely to break even by 2030, productivity gains would have to reach an extraordinary level.

This story is the latest to join our MIT Technology Review Narrated podcast, released weekly on Spotify and Apple Podcasts. Search for MIT Technology Review Narrated on either platform and follow us for new episodes.

The must-reads

We've scanned the internet for today's most compelling technology stories.

1. Hackers claim theft of data on nearly all FBI employees
The ShinyHunters group says it obtained more than 2 TB of information. (Axios)
+ The trove reportedly includes agents' names, addresses, and phone numbers. (404 Media)
+ ShinyHunters describes the breach as payback for an FBI alert. (Reuters $)
+ Now is a good time for doing crime. (MIT Technology Review)

2. Anthropic and OpenAI both launch lower-cost models
The move comes as cheaper Chinese alternatives intensify competition. (CNBC)
+ Startups have been cutting their dependence on the two labs. (Bloomberg $)
+ These are the first new models since both companies urged a slowdown in AI development. (FT $)
+ Their CEOs are scheduled to brief the UN Security Council today. (Quartz)
+ Could AI really kill us all? (MIT Technology Review)

3. Trump's Treasury chief may become his next AI czar
Scott Bessent has been a key figure in US-China AI negotiations. (Semafor)
+ Other candidates include Michael Kratsios and Scott Kupor. (Gizmodo)
+ Will Trump's effort to rebrand AI as "superintelligence" stick? (BBC)

4. Scientists are converting renewable electricity directly into food
The aim is producing more food while using less land. (New Scientist $)
+ Companies are literally making food out of thin air. (MIT Technology Review)

5. A "fire amoeba" shatters the heat tolerance record for complex life
The newly identified organism can grow and reproduce at 63°C. (NPR)
+ The find could shape the search for life beyond Earth. (New Scientist $)

6. Meta quietly trialed a "human concierge" behind its Muse AI agent
Contractors covertly handled some of Muse's calls. (Reuters $)
+ Employees flagged worries about privacy and deceiving users. (404 Media)

7. Data centers are swapping copper for light to reduce energy consumption
Photonics transfers data while generating less heat than electrical wiring. (BBC)
+ Virtual power plants may lend a hand too. (MIT Technology Review)

8. AI models grown from rat brains edge closer to reality
AWS is using rat neurons to speed up video AI. (Wired $)

9. Uber is counting on human drivers to boost its robotaxi strategy
People can absorb demand surges while autonomous vehicles recharge. (Axios)

10. A humanoid robot stepped into a cage against an amateur fighter
The short match was promoted as the first human-robot MMA bout. (Futurism)

Quote of the day

"The use of the word artificial makes it sound fake. It's not fake; it's actually amazing."

—President Donald Trump explains to the UN General Assembly why he prefers to call AI "superintelligence."

One more thing: AI is changing how we quantify pain

At Orchard Care Homes, nurses once assessed pain in residents unable to speak using an observational scale. Agitated residents were sometimes misread as having behavioral problems, leaving their pain untreated.

In 2021, the care-home chain began testing PainChek, a smartphone app that analyzes a resident's face for tiny muscle movements and applies AI to generate a pain score. Within weeks, the pilot unit recorded fewer prescriptions and noticeably calmer corridors.

Researchers are now working to make pain something a camera or sensor can measure as reliably as blood pressure. But when algorithms grade our suffering, does that alter how we understand and treat it?

Explore how AI is reshaping the way we assess pain.

—Deena Mousa

We can still have nice things

A corner for comfort, fun, and distraction to lift your day. (Suggestions? Drop me a line.)

+ Scientists have identified the first new penguin species in over a century.
+ Learn the filmmaking techniques that turned Paul Verhoeven's RoboCop into a directing masterclass.
+ What happens when fast food chains adapt to local palates? These unusual international menu items offer some delicious answers.
+ A media player lets you dance through every track the crowd has IDed at Berghain since 2024, replaying each night in order.