AI News Roundup – Apple chooses Google Gemini AI to power updated Siri, AI tool discovers treatment for rare disease, US imposes new tariffs on AI chip sales to China, and more

To help you stay on top of the latest news, our AI practice group has compiled a roundup of the developments we are following.

• • Apple has chosen Google’s Gemini AI models to power its Siri voice assistant, according to CNBC. In a joint statement between the companies, Apple “determined that Google’s Al technology provides the most capable foundation for Apple Foundation Models and is excited about the innovative new experiences it will unlock for Apple users.” It had been reported for several months in 2025 that Apple was searching for an AI partner as its own efforts to develop an AI-powered Siri were delayed due to several engineering setbacks. The financial details of the AI deal were not disclosed, but Bloomberg reported in November 2025 that Apple was planning to pay around $1 billion per year to use Google’s AI models. Apple currently partners with OpenAI’s ChatGPT for its Apple Intelligence software, though Apple said that its arrangement with OpenAI would not be affected by the new deal with Google. AI-powered updates to Siri and other Apple software are expected later this year.
  • New York Magazine’s Intelligencer reports that researchers at the Mayo Clinic have developed an AI tool that could help develop treatments for rare diseases. Jorie Kraus, a prematurely born baby, was diagnosed with several health issues after birth before being diagnosed with DeSanti-Shinawi syndrome, an extremely rare genetic disorder that affects brain development and is thought to be incurable. However, some members of Jorie’s treatment team have worked for over a year on research that could help children like her. The researchers developed a custom AI tool dubbed BabyFORce that can identify existing drugs that may be able to treat conditions that do not have any known treatments. The AI tool takes in an affected gene and determines whether the activity of that gene must be increased or decreased. For Jorie, the output of the WAC gene needed to be increased, and the researchers identified clonazepam, a generic seizure drug, as promising. After only a few days on a liquid version of the drug, Jorie’s behavior improved drastically, and a later blood test showed that the WAC gene was functioning normally. The researchers are now looking into a clinical trial for DeSanti-Shinawi syndrome using the drug. While the application of AI to medicine has been controversial, research tools like BabyFORce have the potential to help treat previously untreatable conditions.
  • Trump administration has imposed new tariffs on AI chip sales made in China by U.S. semiconductor companies Nvidia and AMD, according to the Financial Times. In a complex scheme, the U.S. government will take a 25% cut of sales of AI chips that are imported into the U.S. but then exported to China. This AI Roundup previously reported on the planned scheme in August 2025, though the planned rate at that time was only 15%. AI chip exports to China have been controversial, especially as the U.S. attempts to retain its lead in the global AI hardware race, but President Trump said in the Oval Office that “I think it’s a very good deal.” The new tariff applies to Nvidia’s H200 and AMD’s MI325X chips, among others, both manufactured by Taiwan’s TSMC, the global leader in semiconductor manufacturing. Such tariffs are made under Section 232 of the Trade Expansion Act of 1962, differing from many of Trump’s other tariffs made under an emergency powers law that is facing a legal challenge at the Supreme Court. Nvidia said this past week that the new policy “strikes a thoughtful balance that is great for America.”
  • Ars Technica reports on new API licensing deals struck between Wikipedia and several AI developers. This past week, the Wikimedia Foundation, the nonprofit which runs Wikipedia, announced API access deals with several AI developers, including Microsoft, Meta, Amazon, and Perplexity, allowing those companies to gain high-volume access to Wikipedia content which is then used for AI training. The new deals mean the companies join Google and several smaller developers in the Wikimedia Enterprise program, which was set up to combat web-scraping of Wikipedia content that was placing strain on the site’s web infrastructure. While Wikipedia’s content remains freely available, the Enterprise program provides faster, higher-volume access to that content which is often a boon to AI developers. The financial terms of the new agreements were not disclosed.
  • The U.S. Senate has unanimously passed a bill aimed to crack down on nonconsensual AI “deepfakes,” according to The Verge. The Disrupt Explicit Forged Images and Non-Consensual Edits (DEFIANCE) Act would allow those who have their likeness “deepfaked” into sexually explicit images to sue the creators of such deepfakes for civil damages. The bill passes as Elon Musk’s xAI has faced growing backlash over its Grok AI model generating sexualized deepfakes of women and girls, including the mother of one of Musk’s children, who has sued the company. As this AI Roundup reported last week, the company is facing investigations from regulators around the world. California and Japan have also recently threatened action against the company. The DEFIANCE Act passed the Senate unanimously and now proceeds to the House of Representatives. The same bill passed the Senate in 2024, though the bill later stalled in the House and never reached the desk of the president. House leadership and President Trump have yet to comment on whether they support the bill.