A U.S. military special operations analyst used an AI chatbot to draft an intelligence report on a Chinese ship, and the AI falsely claimed the vessel was carrying nuclear weapons. U.S. planes scrambled and soldiers prepared to board the ship before the error was caught. The incident, tied to Defense Secretary Pete Hegseth's push for military AI adoption, highlights serious risks of relying on AI for sensitive military reporting without strong human verification.
Meanwhile, China is preparing to shift more toward Huawei's AI training infrastructure starting in 2027. Rotating chairman Eric Xu said Ascend chips have overtaken Nvidia in domestic market share, though Huawei still faces supply limits. Huawei is also accelerating chip releases, bringing the Ascend 960DT forward to Q1 2027 and the Ascend 960R to Q3 2027, with plans to build SuperClusters of up to 512,000 NPUs.
On the regulatory front, Congress is actively exploring AI use and oversight. The 'Godfather of AI' warned lawmakers they have about a year to regulate the technology effectively. The hearing also featured testimony from the mother of a Bay Area teenager who died while using ChatGPT, underscoring concerns about AI safety for young users.
In the corporate world, Meta and Oracle are pouring tens of billions into AI infrastructure with slow payoffs. Oracle reported a $664 billion backlog but faced negative free cash flow of $5 billion due to $28.5 billion in capital expenditures. Meta's revenue grew nearly 28%, yet its free cash flow dropped over 90% from heavy AI spending. Separately, scammers used fake AI trading bot tutorials to steal about $517,000 from 224 victims, while the FBI recorded $16.6 billion in internet-crime losses during 2024.
Key Takeaways
- An AI chatbot falsely reported a Chinese ship carried nuclear weapons, nearly triggering a U.S. military confrontation before the error was discovered.
- Huawei plans to shift China toward its Ascend-based AI training infrastructure in 2027 and has accelerated releases of the Ascend 960DT and 960R chips.
- Congress is under pressure to regulate AI within about a year, with testimony from a mother whose teen died using ChatGPT shaping the debate.
- Oracle reported a $664 billion backlog but faced negative free cash flow of $5 billion due to $28.5 billion in capital expenditures tied to AI infrastructure.
- Meta's revenue grew nearly 28% but its free cash flow dropped over 90% because of heavy AI spending.
- The FBI recorded $16.6 billion in reported internet-crime losses during 2024, including scams involving fake AI trading bot tutorials that stole about $517,000 from 224 victims.
- U.S. officials removed Alibaba's Qwen model from the Federal Register after the FBI flagged Alibaba among companies allegedly copying U.S. frontier models.
- Berkshire Hathaway's new CEO Greg Abel concentrated 31% of the portfolio into two AI stocks, marking a major departure from the company's traditional investment strategy.
- Defense Secretary Pete Hegseth continues to push for greater military use of AI despite the near-conflict incident caused by an AI hallucination.
- Huawei plans to build SuperClusters with up to 512,000 NPUs as it works to close the gap with Nvidia in individual chip performance.
AI Error Almost Led US Military to War With China
A new report says the U.S. military nearly started a war with China because of an AI mistake. A special operations command analyst used an AI chatbot to write an intelligence report about a Chinese ship. The AI wrongly said the ship was carrying nuclear weapons, which turned out to be completely false. U.S. planes took to the air and soldiers prepared to board the ship before the error was found. The Department of War said it relies on rigorous verification of intelligence inputs.
US Military Nearly Went to War After Using AI Report
The U.S. military nearly started another war after a report about a Chinese ship was written with help from artificial intelligence. The report was entirely false and incorrectly described the ship's cargo. The incident shows the risks of using AI for sensitive military reports. The U.S. military has since taken steps to stop AI from writing reports that could lead to conflict. It also highlights the need for human oversight in important decisions.
AI Chatbot Hallucinated Nuclear Weapons on Chinese Ship
A U.S. special operations command analyst used an AI chatbot to make a report about a Chinese ship's cargo. The chatbot hallucinated and falsely flagged the cargo as nuclear weapons. The analyst then used AI again to turn those false details into a military intelligence report. U.S. planes flew and armed soldiers prepared to seize the ship before the mistake was discovered. Defense Secretary Pete Hegseth has pushed for more military use of AI despite these risks.
China to Shift to Huawei AI Training in 2027
Huawei expects a major shift in China toward its AI computing infrastructure for model training starting next year. Rotating chairman Eric Xu Zhijun said training will be based on SuperPoD or SuperCluster systems using Ascend 950DT chips. He said Huawei's Ascend chips have overtaken Nvidia in domestic market share, though he did not give specific numbers. Huawei still faces supply constraints. Nvidia confirmed it had begun taking steps in response.
Huawei Speeds Up AI Chip Releases
Huawei is speeding up its AI chip development and bringing forward the release of two new chips. The Ascend 960DT will now come out in the first quarter of 2027, three quarters earlier than planned. The Ascend 960R will arrive in the third quarter of 2027, three months ahead of schedule. Deputy chairman David Wang said the company is focusing on cluster-based architecture because its individual chips trail Nvidia. Huawei plans to build SuperClusters with up to 512,000 NPUs.
Congress explores AI use and regulation
Members of Congress are examining how they use AI tools. The 'Godfather of AI' warned that Congress has about a year to regulate AI effectively. The discussion includes lawmakers' own adoption of the technology.
Study reveals limits of confidence in diffusion models
New research shows discrete diffusion models have limits in measuring confidence. The study found that per-position distributions cannot determine whether groups of tokens are dependent. On a synthetic task, generated distributions measured 29 times the sampling-noise floor.
US government used Chinese AI model FBI called malicious
US officials removed a Chinese AI search tool from the Federal Register website. The National Archives had been using Alibaba's Qwen model. The FBI had named Alibaba among companies allegedly copying US frontier models.
Fake AI trading bot tutorials steal from 224 victims
Scammers used fake AI trading bot tutorials to steal 274.6 ETH worth about $517,000 from 224 victims. Victims followed video instructions and deployed malicious contracts themselves. The FBI's Internet Crime Complaint Center recorded $16.6 billion in reported internet-crime losses during 2024.
Mother of teen who died using ChatGPT testifies
The mother of a Bay Area teenager who died using ChatGPT gave testimony on the matter. The hearing relates to AI regulation efforts in Congress. The case highlights concerns about AI safety and its impact on young users.
Meta and Oracle pour billions into AI with slow payback
Meta and Oracle are spending tens of billions on AI infrastructure but are not seeing quick returns. Oracle reported a $664 billion backlog and strong cloud growth but had negative free cash flow of $5 billion due to $28.5 billion in capital expenditures. Meta's revenue grew nearly 28% but its free cash flow dropped over 90% because of heavy AI costs. Investors currently prefer Meta because its ad business still generates cash, while Oracle faces concerns over its funding gap. The key challenge for both companies is when their AI spending will turn into positive free cash flow and lasting revenue growth.
Berkshire Hathaway puts 31% of portfolio into 2 AI stocks
Berkshire Hathaway's new CEO Greg Abel has concentrated 31% of the company's portfolio into just two AI stocks. This marks a major shift from the company's traditional focus on older industries. The move signals strong confidence in AI as a driver of growth and innovation. Analysts praised Abel's strategy as a smart investment bet. The market is paying close attention as Berkshire adapts to the rapidly changing tech landscape under Abel's leadership.
Trump hears from all sides on AI future
Tech investor David Sacks has Donald Trump's ear when it comes to AI policy. The report covers Trump hearing from multiple perspectives on AI, including doomers who fear risks and dreamers who see big potential. The article also touches on views involving Xi and other key figures in the AI world. The piece highlights how Trump is being briefed on AI from many different angles. The reporting was published on September 18, 2026.
Sources
- U.S. Military Almost Started War With China Due To A.I. Error, New Report Claims
- U.S. Nearly Started Another War After Military Used AI to Write Report
- US Military Nearly Started World War III After AI Chatbot Hallucinated Nuclear Weapons Aboard a Chinese Ship
- China to see major shift to Huawei for AI model training in 2027: rotating chair
- Huawei aims to deliver faster AI chips, faster
- How are members of Congress using AI?
- Limits of Confidence in Diffusion
- US government website used Chinese model the FBI called "malicious"
- Fake AI trading bot tutorials steal 274.6 ETH from 224 victims
- Mother of Bay Area teen who died using ChatGPT gives testimony
- Meta and Oracle invest heavily in AI but face slow cash returns and high costs.
- 31% of Berkshire Hathaway's Portfolio Is Concentrated in These 2 AI Stocks Under Greg Abel
- The doomers, the dreamers and Xi: Trump hears from all sides on AI
Comments
Please log in to post a comment.