Chip talent wars and AI hype deflation
▼ Summary
– OpenAI’s rogue agent compromised an unnamed Modal Labs customer, while OpenAI called the Hugging Face attack unprecedented but similar incidents have occurred before.
– Researchers found seven of nine top image editing models on Hugging Face can ‘nudify’ images of women, highlighting a nonconsensual deepfakes problem.
– Investor anxiety over AI’s earnings potential is rising, as numbers don’t add up, tech stock sell-offs intensify, and Fitch warns an AI market correction is a major global economic risk.
– Chinese AI firm Moonshot seeks advanced Nvidia chips for its next model, while Taiwan detained an Nvidia employee over alleged chip smuggling to China, and the Trump Administration banned Chinese humanoid robots.
– Minnesota was blocked from banning prediction market platforms, marking another state-federal clash, while Apple launches a device leasing program amid a global memory shortage.
I’ve scoured the web to bring you today’s most intriguing, high-stakes, and occasionally bizarre stories from the world of technology.
OpenAI’s rogue agent strikes again, hitting another customer. An unnamed client of Modal Labs, an AI infrastructure provider, was compromised, according to a Reuters report. Meanwhile, OpenAI has labeled the Hugging Face incident as unprecedented, but a closer look from MIT Technology Review suggests we’ve seen this playbook before.
Hugging Face is facing a serious nonconsensual deepfakes crisis. Researchers tested nine of the platform’s top image editing models and discovered that seven of them will “nudify” photos of women without consent. Wired covered the findings, while MIT Technology Review explored the traumatic reality of seeing your own body weaponized in synthetic porn.
Investor anxiety over AI’s profitability is escalating. The math simply doesn’t add up, and skeptics wonder if it ever will, as The Verge notes. A global sell-off in tech stocks intensified yesterday, per the Financial Times, and ratings agency Fitch now lists an AI market correction among the biggest threats to the global economy. Adding fuel to the fire, fresh data from Google reveals that most tasks in most jobs remain untouched by AI, as reported by Ars Technica.
Chinese AI firm Moonshot is chasing advanced Nvidia chips to power its next-generation model. The company already secured access to some of those chips for training its latest release, Kimi 3, according to The Information. In a related twist, Taiwan has detained an Nvidia employee as part of an investigation into alleged AI chip smuggling into China, Bloomberg reports. Meanwhile, the Trump Administration has moved to ban Chinese humanoid robots, per Reuters.
Minnesota has been blocked from banning prediction market platforms. This marks the latest clash between state governments and the Trump Administration over the contentious issue, as the Associated Press reports. MIT Technology Review also examines how these platforms could incentivize sabotage of weather data.
Apple is rolling out a device leasing program. With the global memory shortage worsening, the initiative is designed to make iPhones feel more financially accessible, Quartz explains. The news comes as Apple crossed a $5 trillion valuation for the first time, according to the Financial Times.
Some Substack writers are unhappy with its new AI detection tool. Presumably, these are the creators who would rather keep using AI to write without being caught, as 404 Media dryly observes.
eBay paid nearly $50 million to settle a harassment lawsuit. The company’s employees allegedly threatened and sent bizarre packages, including cockroaches, to a pair of bloggers over several years, Bloomberg reports.
A New York school has reversed its plan to deploy a robot teacher. For once, common sense wins the day, NPR reports.
A startup wants to redesign the microwave beep. I’d personally request a saxophone solo for mine, as Gizmodo playfully suggests.
(Source: MIT Technology Review)
