BREAKING NEWS: THE FBI and western intelligence agencies had a long history of ordering pagers and/ or walkie-talkies from Gold Apollo of Taiwan, it was revealed yesterday.
The company in New Taipei City made customized communications devices for western intelligence groups for years, according to an unearthed 13-year-old business report.
Before this revelation, the search for the makers of the pager and walkie-talkie bombs used in a devastating Israeli attack on people in Lebanon had moved to a pair of intelligence-linked companies in Europe – but both turned out to be shell companies.
Now, a re-discovered 2011 article in the Chinese language CommonWealth Magazine switches the focus back to Taiwan.
GLOBAL OUTRAGE
The 2011 business profile says that pagers and walkie-talkies may seem "to be antiques", but were actually still being made in large numbers by manufacturer Gold Apollo on the island of Taiwan.
After global outrage over the planting of bombs in pagers and walkie-talkies last week, Gold Apollo staff denied complicity and said they had licenced one of their brands to a western company, BAC Consulting of Hungary.
But that group had no factory, and turned out to be an Israeli intelligence-run front company which also dealt with a Bulgarian firm called Norta Global, the New York Times reported.
Now, the unearthing of the 2011 report moves the discussion back to New Taipei City in Taiwan.
CUSTOMIZED WALKIE-TALKIES
"Gold Apollo sells a walkie-talkie that is almost extinct in Taiwan, but its professional technology, customization and strict quality control have won orders from European and American governments, making it number one in the United States and the second largest in Europe," the 2011 CommonWealth Magazine report said.
The company in Taiwan specializes in pagers and "customized walkie-talkies", the article continued. Customers include intelligence groups in various countries, the report said. "The demand is mainly concentrated in intelligence, firefighting, national defense and other units in Europe and the United States."
The US Federal Bureau of Investigation is a customer. "The FBI ordered a text-type walkie-talkie," said the report, based on interviews with the company's staff. It says of the FBI order: "The technical requirements of this product are not high, but the security confidentiality requirements are extremely strict."
The small company outsources key parts of the production chain to others. Gold Apollo "is responsible for receiving orders, and designing and purchasing raw materials", and also "production and assembly outsourcing", the article added.
GOVERNMENT FOUND 'NO RECORDS'
Taiwan's Ministry of Economy last week said they found no records indicating that the company had directly exported goods to Lebanon. But the statement said nothing about indirect supply chains, which is how most goods move from the island.
Taiwan is legally recognized as part of China, but is presently managed by an increasingly unpopular Washington-aligned political party.
Senior staff at Gold Apollo last week portrayed the company as a victim of a scam and threatened to take out a lawsuit.
But the revelation that there is a long history of people in Taiwan making customized pagers and walkie-talkies for western intelligence groups will potentially change the narrative.
by Nury Vittachi
Lai See(利是)
** 博客文章文責自負,不代表本公司立場 **
Washington is threatening to sanction Chinese AI models. Nvidia CEO Jensen Huang just told Axios something very different. In an exclusive interview in Fort Worth, Texas on July 21, Huang said US companies should be "absolutely" allowed to use open-source Chinese AI models. He went further: the odds that Chinese firms drive American companies out of business are zero.
Huang's answer was one word: "Absolutely." US firms should be allowed to use Chinese AI.
Huang's comments read like a direct warning to hawks in Washington pushing for a ban. The trigger was Moonshot AI's July 16 release of Kimi K3, a 2.8-trillion-parameter model and now the world's largest open-source system. Its aggressive pricing and upcoming open-weight downloads set off the sharpest wave of AI panic since DeepSeek rattled markets in early 2025. Kimi K3 beat Anthropic's Claude Opus 4.8 on multiple benchmarks, and chip stocks including Nvidia sold off on fears that cheaper, more efficient models would undercut the massive global bet on AI infrastructure.
The fallout came fast. US Treasury Secretary Bessent and Trade Representative Greer both spoke out the same day. Bessent told Fox Business that US authorities had found watermarks from American large language models embedded in several Chinese AI systems, calling it "unacceptable" and promising an investigation. Greer said Washington would examine whether China gained its edge through unfair means.
Bessent: “US watermarks found in Chinese AI models.” An investigation is coming
US media reported earlier that the White House is weighing a new executive order to restrict open-source AI, particularly systems from Chinese companies. The Commerce Department had already considered adding several Chinese AI labs to its Entity List last year.
Huang's logic cuts the other way. "The market misunderstood the impact of DeepSeek the first time," he told Axios, adding that Wall Street has "misunderstood the impact of Kimi again this time."
Kimi K3's launch rattled the entire AI industry.
Free, cheap open-source models, according to Huang, let AI reach far more consumers and businesses. That, he argued, drives strong demand for chips, data centers and computing power. "Free AI should be great for hardware," Huang said. "Free AI should be great for chips. Free AI should be great for data centers."
OpenAI and Anthropic have no reason to fear open models either, in Huang's view. Open-source systems expose more people to AI for the first time, which expands the whole market. Many of those users, he added, will still end up paying for the reliability and convenience of closed commercial services.
On security, Huang rejected the idea that downloading a Chinese model opens a backdoor to Beijing. Companies can fully customize and control access inside secure sandbox environments, he said.
Wong argued openness actually works the other way: outside researchers can inspect models, spot weaknesses and build defenses. "If everything just becomes one single model, one single point of attack, one single source of failure, I think the world is much, much more vulnerable."
Huang also called on Anthropic to open up its restricted cybersecurity model, Claude Mythos, rather than keep it locked down. "Holding Anthropic back is not in the benefit of the United States," he said, "Open models are available anyhow. So I think: Let Anthropic run."
On the US-China AI race, Huang staked out a position sharply at odds with Washington's. He rejected the idea that AI competition has a finish line. "We're going to continue to use AI forever," he said. "The United States is going to be here for a long time. China's going to be here for a long time."
Bloomberg cited several analysts drawing a contrast with the 2018 ban on Chinese 5G equipment. Freely downloadable, highly customizable software models have already embedded themselves deep in Silicon Valley's ecosystem. On OpenRouter, a platform offering access to multiple AI models, Chinese models now account for nearly 60% of token usage by US companies.
The adoption numbers back that up. DoorDash, Airbnb and Germany's Siemens have all shifted to Chinese models. More than 80% of US AI startups now make heavy use of Chinese models such as DeepSeek and Zhipu's GLM, with usage share holding above 30%, up from just 11% a year ago.
Nathan Lambert, an AI researcher at the University of California, Berkeley, put it bluntly: any form of ban would be a serious mistake for AI's long-term development.
Morgan Stanley believes the growth cycle in global AI computing demand is far from over. Bernstein estimates that even if Kimi K3 cuts single-run training costs by roughly 40%, competitive pressure pushing firms to double their training frequency could still lift total computing demand by around 20%.
