Semiconductor giant Nvidia is facing unexpected new U.S. export controls on its H20 chips.

In a filing Tuesday, Nvidia said it was informed by the U.S. government that it will need a license to export its H20 AI chips to China. This license will be required indefinitely, according to the filing — the U.S. government cited “risk that the [H20] may be used in […] a supercomputer in China.”

Nvidia anticipates $5.5 billion in related charges in its Q1 2026 fiscal year, which ends April 27. The company’s stock was down around 6% in extended trading.

The H20 is the most advanced AI chip Nvidia can export to China under the U.S.’ current and previous export rules. Last week, NPR reported that CEO Jensen Huang might have talked his way out of new H20 restrictions during a dinner at President Donald Trump’s Mar-a-Lago resort, in part by committing that Nvidia would invest in AI data centers in the U.S.

Perhaps not-so-coincidentally, Nvidia announced on Monday that it would spend hundreds of millions of dollars over the next four years manufacturing some AI chips in the U.S. Pundits were quick to point out that the company’s commitment was light on the details.

Multiple government officials had been calling for stronger export controls on the H20 because the chip was allegedly used to train models from China-based AI startup DeepSeek, including the R1 “reasoning” model that threw the U.S. AI market for a loop in January.

Nvidia declined to comment.

In the wake of criticism over the underwhelming performance of its AI products, especially in areas like notification summaries, Apple on Monday detailed how it is trying to improve its AI models by analyzing user data privately with the aid of synthetic data.

Using an approach called “differential privacy,” the company said it would first generate synthetic data and then poll users’ devices (provided they’ve opted-in to share device analytics with Apple) with snippets of the generated synthetic data to compare how accurate its models are, and subsequently improve them.

“Synthetic data are created to mimic the format and important properties of user data, but do not contain any actual user generated content,” the company wrote in a blog post. “To curate a representative set of synthetic emails, we start by creating a large set of synthetic messages on a variety of topics […] We then derive a representation, called an embedding, of each synthetic message that captures some of the key dimensions of the message like language, topic, and length.”

The company said these embeddings are then sent to a small number of user devices that have opted in to Device Analytics, and the devices then compare them with a sample of emails to tell Apple which embeddings are most accurate.

The company said it is using this approach to improve its Genmoji models, and would in the future use synthetic data for Image Playground, Image Wand, Memories Creation and Writing Tools as well as Visual Intelligence. Apple said it would also poll users who opt in to share device analytics with synthetic data to improve email summaries.

HR tech startup Rippling is trying to serve Deel CEO Alex Bouaziz with papers as part of a blockbuster lawsuit against its competitor. There’s just one problem: French bailiffs hired by Rippling can’t seem to find Bouaziz, Irish newspaper Business Post reports.

Rippling sued Deel earlier this year in Ireland alleging the company paid one of its employees to spy on them, according to the lawsuit that includes an affidavit from the alleged spy which reads like a movie. Deel denies all wrongdoing.

Rippling needs to serve Bouaziz papers for the case. But French bailiffs hired by Rippling have been unable to find Bouaziz, who is French, at his listed address in Paris. Rippling may need to apply to serve him via email, the Business Post reports.

Bouaziz is “highly nomadic” but remains “very attached to France,” French magazine Challenges reported. Bouaziz lists his location as Tel Aviv on LinkedIn, but New York on X. 

Deel, Rippling, and Bouaziz didn’t respond to requests for comment.