News-July-28-2026

July 28, 2026 — Taiwanese prosecutors detained an Nvidia employee in a widening investigation into the alleged diversion of restricted AI servers to China; the UK Court of Appeal reversed a stay in a criminal trial involving AI-assisted witness preparation and urged justice-system bodies to regulate AI use in criminal proceedings; the U.S. Federal Communications Commission blocked authorization of new foreign-made advanced robots and connected power inverters on national-security grounds; more than 1,000 employees of leading AI companies asked the U.S. government to support an international effort to develop tools for pacing frontier AI development; and AI-infrastructure startup Runlayer sued Rippling over alleged trade-secret misappropriation and product cloning. Separately, the Illinois Appellate Court for the First District imposed a $15,000 sanction after finding ten false quotations or citations in briefs whose attorney acknowledged using ChatGPT.

Contents

1. Taiwan detains Nvidia employee in AI-server smuggling probe
2. UK Court of Appeal orders retrial and urges AI rules for criminal proceedings
3. FCC blocks new foreign-made advanced robots and connected power inverters
4. AI-lab employees urge U.S.-backed international pacing tools
5. Runlayer sues Rippling over alleged MCP gateway cloning
6. Illinois appellate court imposes $15,000 sanction for AI-related citation errors

---

Taiwan detains Nvidia employee in AI-server smuggling probe

The Keelung District Prosecutors Office detained an Nvidia employee on July 28 as part of an investigation into whether restricted AI servers were routed to China. Local media identified the employee as a senior business-development manager, and The Next Web reported that he had not been charged and that the allegations had not been tested in court.

Investigators searched the employee's home and desk at Nvidia's Taipei office on July 24, after which prosecutors sought and obtained court permission to detain him on allegations of forgery and breach of trust. Taiwanese press accounts cited by The Next Web allege that he approved end-user and know-your-customer documents used during export review before servers were diverted to China.

The detention extends an investigation that had previously centered on Super Micro Computer and Taiwanese distribution and data-center partners. Nvidia had not publicly commented on its employee's detention when the reports were published.

---

UK Court of Appeal orders retrial and urges AI rules for criminal proceedings

In Rex v FGD, the Court of Appeal allowed a prosecution appeal against a recorder's decision to stay a rape trial after the complainant was found to have used AI to help prepare for court, and the appeal court ordered a retrial at the earliest opportunity.

The judges said the case was unlikely to be isolated and invited the National Police Chiefs' Council, Crown Prosecution Service, and Criminal Procedure Rule Committee to consider how to address and regulate AI use in criminal proceedings. The court distinguished prohibited witness training from the circumstances before it, noting that the AI-generated coaching material was recorded and could be placed before the jury, and concluded that the trial process could manage any resulting prejudice.

---

FCC blocks new foreign-made advanced robots and connected power inverters

On July 28, the Federal Communications Commission added foreign-produced mobile robots and connected power inverters to its Covered List, preventing new models from receiving the equipment authorization generally required before electronic devices can be imported, marketed, or sold in the United States.

The restriction covers advanced robotic devices including humanoid and quadruped robots and applies to products made in any foreign country, while administration officials described the action as protecting the U.S. AI supply chain from disruption, data theft, and cyberattacks associated with Chinese technology.

The FCC's national-security findings cited risks arising from remotely connected equipment, including manipulation of robots' data or physical operation and the possibility that inverter access could be used to disrupt energy systems, remove data, conduct surveillance, or enable cyberattacks.

Previously authorized models may continue to be imported, marketed, sold, and used, while manufacturers of new models may seek conditional approval from the Department of Defense for robots or from the Defense or Homeland Security departments for power inverters.

---

AI-lab employees urge U.S.-backed international pacing tools

More than 1,000 employees at leading U.S. AI companies signed the July 28 “Pacing the Frontier” statement, which asks the U.S. government to support an international effort to develop technical and governance tools capable of deliberately pacing frontier automated-AI development.

The statement says industry, government, and society may need an option to buy time to address emerging risks, develop security measures, and strengthen oversight; it does not demand an immediate slowdown or pause.

NBC News reported that the signatories included the chief scientists or science officers of Anthropic, OpenAI, and Meta, along with senior leaders from Google and Thinking Machines.

---

Runlayer sues Rippling over alleged MCP gateway cloning

Runlayer, which sells a security gateway for connecting AI models and agents to outside data and tools through the Model Context Protocol, filed suit against HR-software company Rippling after a prospective-customer product trial. Runlayer alleges that the parties signed a mutual nondisclosure agreement and a trial agreement barring copying or derivative works, and that the nearly year-long evaluation included access to Runlayer's roadmap and source code.

The complaint alleges trade-secret misappropriation, unfair competition, and breach of contract, asserting that Rippling later developed a substantially similar MCP gateway. Rippling confirmed that it was launching its own gateway but denied misusing Runlayer's intellectual property.

---

Illinois appellate court imposes $15,000 sanction for AI-related citation errors

In a July 28 opinion in Scott v. Illinois Human Rights Commission, the Illinois Appellate Court for the First District ordered attorney Mason Cole to pay $15,000 after finding that briefs he filed contained ten false statutory quotations or case citations as well as improper statements of fact.

The court assessed $1,500 for each false citation or quotation and said courts may need to increase fines until sanctions meaningfully deter AI-hallucinated citations. Cole acknowledged that he frequently used a corporate ChatGPT subscription to synthesize complex matters but said he tried to cross-check citations in LexisNexis.

The decision affirmed an Illinois Human Rights Commission order sustaining the dismissal of Starbucks employee Kimberly Scott’s workplace racial-discrimination charge.