Welcome to the newest edition of ID Tech’s AI Update. Here’s the latest big news on the shifting landscape of AI and identity technology:
Taiwan Semiconductor Manufacturing Company (TSMC) is suing former SVP Wei-Jen Lo over trade secrets concerns. Lo joined Intel last month to help with its chip foundry efforts after about 21 years with TSMC, and the latter says he is breaching a non-compete clause in his contract and violating Taiwan’s Trade Secrets Act. TSMC argues there is a “high risk” that Lo could use or disclose its trade secrets at Intel, given his deep knowledge of cutting-edge process technologies. Intel says TSMC’s allegations are baseless and without merit.
Major Chinese firms like Alibaba and ByteDance are using third-country data centers to bypass US chip restrictions, according to a new Financial Times report. The firms are alleged to be training their AI models using Nvidia chips by renting access to compute through countries like Malaysia and Singapore. US sanctions prohibit the sale of Nvidia chips to China, but not to other friendly Southeast Asian countries.
Alibaba has started selling AI-powered smart glasses in China, marking a rare foray into consumer hardware for the tech giant. The Quark AI Glasses come in two variations, with the premium S1 version featuring clear micro-OLED displays that overlay information directly in the user’s field of view. Both the S1 and the more budget-friendly G1 feature a voice-based AI assistant based on Alibaba’s Qwen AI model.
China-based VC firm Monolith Management has raised about $289 million for domestic AI investments, becoming one of just a handful of Chinese firms to raise substantial sums of US dollars. Monolith is best known as an early backer to Moonshot AI, a rival of the DeepSeek AI platform that has recently elevated its profile with its Kimi AI models. Monolith is focused on “concentrated bets” and fast term-sheet turnaround, and says it’s willing to underwrite technical risk as long as commercialization paths are visible.
A new executive order from the White House has established a program meant to use AI for scientific discovery. The “Genesis Mission” will be led by the Department of Energy, and is meant to establish an integrated AI platform that links federal scientific datasets with national-lab and private supercomputers to train “scientific foundation models” and deploy AI agents that can design experiments, test hypotheses, and help solve problems in energy, national security, health, and other strategic domains.
New research from Foaster, an AI consultancy, has mapped the political leanings of major AI models. They didn’t find any radicals in the bunch: All of the AI models were a little bit more libertarian than authoritarian, and none of the most popular Western models – ChatGPT, Claude, and Grok – approached the fringes of the left-right political spectrum. GPT-5, in fact, was squarely in the center, with Gemini a little to the left, and Grok being the only right-of-center AI model. France’s Magistral was furthest from the center, and also happened to be the least predictive with respect to actual voter preferences. Grok, meanwhile, was most predictive of election results. (Thanks to Tyler Cowen for sharing this.)
The chatbot’s take: A really interesting insight on AI model’s political/policy preferences:
–
By Alex Perala
