Anthropic has accused several Chinese AI labs, including Alibaba, Moonshot AI, and DeepSeek, of secretly routing user requests to Claude to steal training data. Reports indicate that between May and July 2026, Alibaba alone conducted over 151 million unauthorized exchanges, while DeepSeek transferred more than 12 million. These operations involved thousands of fake accounts and exposed sensitive information, ranging from internal codes to surveillance footage.
Seven specific labs, including Zhipu, Xiaomi, SenseTime, and MiniMax, participated in these industrial-scale attacks. Anthropic confirmed that some actors used the stolen data for state-sponsored surveillance targeting dissidents in China, Iran, and West Africa. In response, the company has updated its security measures to prevent further unauthorized access.
Meanwhile, Meta suspended its elite AI unit after workers revolted over a request to use employee data for model training, which led to a data leak. This internal conflict highlights growing concerns about data privacy within major tech firms.
On the regulatory front, Senator Josh Hawley addressed AI impacts during a February 10, 2026, Senate hearing. Lawmakers are debating how to regulate powerful AI risks without stifling innovation, as warnings about potential societal harm increase.
Outside the controversy, NASA and IBM launched the NASA-IBM Lunar Foundation Model, an open-source tool available on Hugging Face. Trained on 2 million image tiles from 17 years of data, it helps researchers map lunar craters and locate ice. Simultaneously, Fidji Simo, formerly of OpenAI and Meta, joined Nscale's board to guide its AI infrastructure growth.
Investors are also watching the AI hardware sector closely. Analysts suggest Micron Technology could see 60 percent upside compared to Nvidia's 48 percent, driven by the massive expansion of data centers. Y Combinator CEO Garry Tan advocates for open-weight model distillation in the US to prevent single-company dominance.
Key Takeaways
- Anthropic accuses Alibaba, Moonshot AI, and DeepSeek of stealing millions of user conversations to train their models.
- Alibaba led the data theft campaign with over 151 million unauthorized exchanges between May and July 2026.
- Seven Chinese labs, including Zhipu and Xiaomi, used fake accounts to bypass Claude's security restrictions.
- Stolen data included sensitive government credentials and surveillance footage used to target dissidents.
- Meta suspended its elite AI unit following a worker revolt over unauthorized use of employee data.
- Senator Josh Hawley testified before Congress on February 10, 2026, regarding AI privacy and technology impacts.
- NASA and IBM released the Lunar Foundation Model on Hugging Face for free public use.
- Fidji Simo, ex-OpenAI CEO, joined Nscale's board to oversee AI infrastructure development.
- Analysts predict Micron Technology could offer 60 percent stock upside versus Nvidia's 48 percent.
- Y Combinator CEO Garry Tan urges US labs to distill frontier models to avoid corporate monopolies.
Anthropic accuses Kimi and DeepSeek of stealing data
Anthropic claims that Moonshot AI and DeepSeek secretly sent user requests to Claude instead of their own models. This practice allowed them to collect millions of conversations to train their systems without user permission. The reports state that sensitive data, including internal codes and surveillance footage, was exposed during these operations. Anthropic says it has now updated its security measures to stop these attacks.
Alibaba and other labs used millions of Claude chats
Anthropic released a report stating that Chinese AI labs secretly used hundreds of millions of Claude interactions to train their own models. Alibaba led the largest campaign with over 151 million exchanges between May and July 2026. Moonshot AI and DeepSeek also routed user requests to Claude to extract reasoning data for their models. Some of the exposed data included sensitive government credentials and internal company documents.
Seven Chinese AI labs ran attacks on Claude models
Anthropic identified seven China-based AI labs that ran industrial-scale attacks to steal data from Claude. The companies involved include Alibaba, Moonshot, DeepSeek, Zhipu, Xiaomi, SenseTime, and MiniMax. These labs used networks of fake accounts to bypass security restrictions and harvest user conversations. One attack by Alibaba alone involved over 3,500 fraudulent accounts and peaked at 3 million daily exchanges.
Anthropic stops AI surveillance targeting dissidents
Anthropic announced it disrupted state-sponsored surveillance operations using its AI models in China, Iran, and West Africa. These campaigns targeted dissident communities and pro-democracy activists between January and July 2026. The company also found that Chinese developers used Claude to generate answers while secretly stealing the data for training. Some cases involved sensitive biological research where the intent of the actors was not fully clear.
Chinese firms secretly used Claude data for training
Anthropic reported that Alibaba and Moonshot used unauthorized methods to extract capabilities from Claude for their own models. The Alibaba operation was the largest ever measured, involving more than 151 million exchanges in a three-month period. Moonshot silently forwarded customer requests to Claude and displayed the answers as if they came from Kimi. DeepSeek also transferred over 12 million exchanges to Claude without notifying its users.
Senator Josh Hawley speaks on AI at Capitol Hill hearing
Senator Josh Hawley from Missouri spoke during a Senate hearing on artificial intelligence. The event took place on February 10, 2026, at the Capitol Hill in Washington. The hearing was held by the Subcommittee on Privacy, Technology, and the Law. Senator Hawley addressed the committee regarding the impact of AI technology.
Senator Josh Hawley speaks on AI at Capitol Hill hearing
Senator Josh Hawley from Missouri spoke during a Senate hearing on artificial intelligence. The event took place on February 10, 2026, at the Capitol Hill in Washington. The hearing was held by the Subcommittee on Privacy, Technology, and the Law. Senator Hawley addressed the committee regarding the impact of AI technology.
NASA and IBM launch new AI model for lunar science
NASA and IBM have launched the NASA-IBM Lunar Foundation Model to help study the Moon. This open-source AI tool uses data from NASA's Lunar Reconnaissance Orbiter to analyze the Moon's surface. The model was trained on roughly 2 million image tiles collected over 17 years. It helps researchers map craters, find volcanic features, and locate ice near the poles. Kevin Murphy, a NASA data officer, said this tool turns large data into new discoveries.
NASA and IBM use new AI to map the Moon in detail
IBM and NASA created the Lunar Foundation Model to map the Moon's surface using decades of data. The AI uses a technique called masked-token learning to connect different types of lunar data. It performs better than other models at finding craters and estimating where ice exists. Researchers can customize the model for specific tasks without high computing costs. The tool is available for free on the Hugging Face repository for scientists to use.
Learnbay updates AI courses as India hiring grows 33 percent
Learnbay has updated its AI curriculum to match the changing job market in India. AI and machine learning job roles in India grew by 33 percent in July 2026. The Naukri JobSpeak Index shows this is the fastest-growing segment in the market. Learnbay now teaches skills like Retrieval-Augmented Generation and AI agent development. The updated courses also cover system engineering and AI governance for enterprise needs.
Meta suspends elite AI unit after worker revolt
Meta CEO Mark Zuckerberg announced the suspension of the company's elite AI program. The unit had requested employee data to train its models, which caused a leak and an internal revolt among workers. The program has now been halted to address these concerns.
Congress faces pressure to regulate powerful AI risks
New warnings about artificial intelligence risks are making it urgent for Congress to act on the issue. Lawmakers are debating how to create safeguards for advanced AI models without stopping innovation. Some experts worry that sophisticated AI could be used in harmful ways for society. While some politicians want stricter rules, others fear regulations might slow down progress. The debate continues with no clear solution in sight.
Fidji Simo joins Nscale board of directors
Nscale announced that Fidji Simo has joined its Board of Directors effective immediately. Simo previously served as CEO of AGI Deployment at OpenAI and led Instacart to profitability and its 2023 public listing. She brings over 20 years of experience building technology products at Meta, eBay, and other major companies. Nscale is a London-based AI infrastructure company that builds data centers and delivers GPU cloud services. Simo noted that future AI success depends on integrating the entire stack from physical infrastructure to software.
Micron stock offers 60 percent upside versus Nvidia
Investors are comparing Nvidia and Micron Technology as popular choices for AI chip stocks. Both stocks have Strong Buy ratings from Wall Street analysts according to TipRanks data. Micron appears cheaper with a PEG ratio of 0.03 compared to Nvidia's 0.46. Analysts suggest Micron could have 60 percent upside potential while Nvidia offers 48 percent. Nvidia commands a higher price due to its AI leadership, while Micron's lower valuation reflects the cyclical nature of memory chips.
Y Combinator CEO wants US labs to distill frontier models
Y Combinator CEO Garry Tan wants American open-weight AI labs to use distillation techniques on frontier models. He believes this approach should remain free from government regulation to create more open-weight options in the US. Distillation involves prompting a model to learn how another model works and reasons. Tan argues that proprietary labs did not ask permission when they trained on vast amounts of public data. He warns that allowing one single company to control all AI power would be a dangerous outcome for the industry.
AI boom growth benefits NVIDIA Micron and SanDisk
The AI industry is expanding rapidly, bringing new revenue opportunities for major chipmakers. NVIDIA, Micron, and SanDisk are key suppliers with multi-year contracts that Wall Street might undervalue. Jensen Huang describes this infrastructure expansion as the largest in human history. These companies provide essential hardware for data centers and autonomous vehicles while also developing new AI software. This growth wave will create jobs and improve quality of life for people worldwide.
Sources
- Kimi And DeepSeek Served Claude To Their Users Instead Of Their Own Models To Collect Exchanges For Model Training, Says Anthropic
- Anthropic says Chinese AI labs secretly used hundreds of millions of Claude exchanges to train their models
- Anthropic Says Seven China-Based AI Labs Ran Industrial-Scale Claude Distillation Attacks
- Anthropic foils AI surveillance targeting dissidents in China, Iran and West Africa
- Chinese AI labs secretly used millions of Claude exchanges to train their models, Anthropic says
- Congress Artificial Intelligence
- Congress Artificial Intelligence
- NASA, IBM Launch AI Foundation Model for Lunar Science
- Moon's surface will be mapped in incredible detail using first-of-its-kind AI, NASA and IBM reveal
- Learnbay Updates AI Curriculum as India’s AI Hiring Increases by 33%, Driven by Enterprise Deployment
- Meta's AI program prompted internal revolt, program suspended
- Risks of artificial intelligence are getting harder for Congress to ignore
- Fidji Simo Joins Nscale Board of Directors
- Nvidia vs. Micron: One AI Chip Stock Is Cheaper with 60% Upside
- Y Combinator's Garry Tan wants U.S. open-weight AI labs to 'distill' frontier models, too
- The Next Phase of the AI Boom Could Be Great News for NVIDIA, Micron, and SanDisk
Comments
Please log in to post a comment.