I’ve been trying to stay off social media these past few days—as much as I can as a journalist. On Thursday, a video popped up on my X feed, displaying the very moment that political activist Charlie Kirk was shot in the neck as he sat on a stool onstage at a college campus, talking with students and observers. There was no warning, no prompt before the video started playing—only an instant replay of the exact moment of gory violence that would immediately make waves around the country and the world. The video was entirely unexpected, and it momentarily paralyzed me, leaving me unsettled and struggling to focus on the rest of the workday.

I wasn’t alone. During class or in the school hallways, teens watched various versions of the videos of Kirk’s killing that circulated online, sometimes in slow motion. Andrew Apsley, an English teacher in Utah, said a video was sent to his 19-year-old child, who has autism and has difficulty processing emotions. It was “pretty traumatic,” he told the Associated Press.

Over the last few weeks—as we have witnessed the tragic killing of Kirk, in close succession to the horrific videos and images of the stabbing of Ukrainian refugee Iryna Zarutska in Charlotte that also went viral—it has felt like I’ve been forced to repeatedly bear witness to graphic and violent content in the exact moments when I am least emotionally prepared to do so. And it’s sobering to realize that this phenomenon is related, in part, to the business decisions of big social-media companies. In a chilling piece in Wired last Thursday, Lauren Goode pointed out that, as these platforms have scaled back their moderation efforts, researchers say they’re “falling short in enforcing their own content moderation rules, at a moment when political tensions and violence are flaring.”

Alex Mahadevan, the director of MediaWise at the Poynter Institute, told Goode that “this is all psychologically damaging to our society in ways we don’t understand yet.” I would agree. These videos have been replaying in my head since I first saw them and have made it difficult at times for me to get through the day. For me, the only absolute solution has been to stay completely offline. Perhaps I am too sensitive in that I can’t handle sudden, unexpected graphic content on my social media feeds without it shaking me to my core. But I doubt I’m alone in that.

Today’s news below.

—Jessica Mathews

Want to send thoughts or suggestions to Fortune Tech? Drop a line here.

China ratchets up trade-talk pressure with Nvidia antitrust ruling

AI chipmaker Nvidia has increasingly found itself caught in a tug-of-war between Washington and Beijing. And on Monday, China gave a mighty pull, ruling that Nvidia had violated the country’s antitrust laws.

China’s State Administration for Market Regulation (SAMR) said, in a preliminary finding, that Nvidia had failed to fully comply with provisions it had imposed on the chipmaker in 2020 when the agency conditionally approved Nvidia’s acquisition of Mellanox Technologies, an Israeli-U.S. networking equipment maker.

The move is largely seen as a way for Chinese officials to signal their displeasure with Washington’s restrictions on the export of cutting-edge technology, including Nvidia’s top-of-the-line AI chips, to China. The regulator began its antitrust probe of Nvidia’s $6.9 billion acquisition of Mellanox in December, just days after the U.S. unveiled tougher export restrictions on high-bandwidth memory chips, which are important for AI applications, as well as chipmaking equipment.—Jeremy Kahn

New studies show how people really use ChatGPT and Claude

Rival AI companies Anthropic and OpenAI have released dueling studies that paint a picture of how people are using their flagship products, ChatGPT and Claude. Both pieces of research analyzed large datasets of user conversations, examining work and non-work-related conversations.

While the two companies used different datasets and methods, OpenAI’s analysis suggests the consumer versions of ChatGPT are used mainly for personal and exploratory purposes, whereas Anthropic’s findings show Claude.ai and Claude API are primarily used for work-related tasks like coding, research, and education.

According to the study released by OpenAI, most ChatGPT conversations aren’t about work at all. Non-work-related messages made up more than 70% of all usage, up from 53% in June 2024, while work-related queries made up 27% of all messages, down from 47% of total conversations a year ago. –Beatrice Nolan

Scott Bessent says Trump and China have the framework of a TikTok deal

A framework deal has been reached between China and the U.S. for the ownership of popular social video platform TikTok, U.S. Treasury Secretary Scott Bessent said after weekend trade talks in Spain.

Bessent said in a press conference after the latest round of trade talks between the world’s top two economies concluded in Madrid that U.S. President Donald Trump and Chinese Premier Xi Jinping would speak on Friday to possibly finalize the deal. He said the objective of the deal would be to switch to U.S. ownership.

“We are not going to talk about the commercial terms of the deal,” Bessent said. “It’s between two private parties. But the commercial terms have been agreed upon.”—The Associated Press

More Tech

—UAE chips, a Trump crypto bonanza, and an increasingly cozy relationship. A New York Times investigation uncovered details about the connections between a lucrative transaction involving the Trump family’s cryptocurrency firm and an agreement giving the Emiratis access to A.I. chips; White House "AI czar" and venture capitalist David Sacks was reportedly a key player.

—Is AI easy? Marc Benioff and some of his customers disagree. Marc Benioff has told customers they could set up AI agents to handle customer service discussions and other tasks in minutes—but Salesforce has struggled to sell Agentforce, in part because of the extensive prep work customers need to do to make it work, according to The Information.

—Squeaky-wheels AI workers get fired in fight over working conditions. Over 200 contractors who work on improving Google’s AI products, including Gemini and AI Overviews, have been laid off, sources told Wired.

—Will 'violent task churn' be as bad as it sounds? Yes, AI will transform and maybe eliminate millions of jobs, but the ensuing productivity boom will generate better ones, says a JPMorgan report.