• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
Technology

Alibaba allegedly created more than 151 million fraudulent Claude interactions to improve Qwen

A user's AI news roundup relays Anthropic's allegations against Alibaba, alongside claims that Moonshot and DeepSeek routed real customer conversations through Claude to collect training data.

GabrielGA
1 Source, 25d ago, first seen 25d ago

TLDR

A user's September 11, 2026, news roundup says Anthropic alleged that Alibaba created more than 151 million fraudulent interactions with Claude in an attempt to improve Qwen. It also relays allegations that Moonshot and DeepSeek routed real customer conversations through Claude to collect training data. The roundup says Anthropic's threat report also details Russian espionage operations and researchers attempting to use Claude for work that could support biological weapons.

Combined views

387

1 Source, first seen 25d ago

2 likes1 saves

Combined views

387

1 Source, first seen 25d ago

2 likes1 saves

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

1 Source

Gabriel@gabriel_horwitzAI NEWS WAS NONSTOP TODAY. Here's a full breakdown: 1. Anthropic says Alibaba created more than 151m fraudulent interactions with Claude in an attempt to improve Qwen, while Moonshot and DeepSeek allegedly routed real customer conversations through Claude to collect training data. Its new threat report also details Russian espionage operations and researchers attempting to use Claude for work that could support biological weapons. 2. China’s AI chipmakers are raising prices as the global HBM shortage gets worse. Huawei increased quotes for its next Ascend accelerator by roughly 20% to 50%, Cambricon raised prices by 20% to 30% and smaller manufacturers are reportedly reserving chips for major customers such as ByteDance. Memory is becoming a serious constraint on China’s Nvidia alternatives. 3. DeepSeek launched V4.1-Flash, the smallest model built on its new V4.1 architecture. The company says it delivers faster inference and higher throughput while providing a foundation for larger models, one day after reportedly hiring an adviser for its Shanghai IPO. 4. The Justice Department is investigating whether Nvidia structured its $17bn Groq licensing agreement to avoid a normal merger review. Nvidia licensed Groq’s chip technology and hired its founder + several executives without technically acquiring the company, which is exactly the type of AI “reverse acquisition” regulators have started questioning. 5. A bipartisan group of senators opened investigations into OpenAI’s handling of the Hugging Face breach and other rogue-agent incidents. Josh Hawley demanded internal documents and a full account from Sam Altman by October 1, while Chris Van Hollen wants federal cybersecurity agencies to receive access to OpenAI’s advanced models. 6. AI chip startup Positron raised $875m at a $5bn valuation, up from roughly $1bn seven months ago. Its upcoming Asimov processor uses a memory-first architecture with as much as 2.3 terabytes of memory, targeting the increasingly massive inference market rather than model training. 7. IBM and NASA released an open-source lunar foundation model trained on data from nine instruments across four NASA missions. It identifies ice deposits, craters and volcanic formations with up to 23% greater accuracy, helping researchers locate landing sites and resources needed for a long-term Moon base. 8. Nvidia signed agreements supporting 2 GW of new Australian AI data-center capacity by 2027. Local companies including Firmus, AirTrunk, NextDC and CDC will operate the facilities while Nvidia supplies the computing, networking and software stack. 9. Latham & Watkins became the first major law firm publicly known to purchase Nvidia servers and build its own internal AI infrastructure. It is customizing open-weight models, leasing secure data-center space and employing more than 900 technology specialists so it can keep sensitive client data away from external cloud providers. JUST ANOTHER NORMAL DAY IN AI.25d
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI

    1 Source

    Gabriel@gabriel_horwitzAI NEWS WAS NONSTOP TODAY. Here's a full breakdown: 1. Anthropic says Alibaba created more than 151m fraudulent interactions with Claude in an attempt to improve Qwen, while Moonshot and DeepSeek allegedly routed real customer conversations through Claude to collect training data. Its new threat report also details Russian espionage operations and researchers attempting to use Claude for work that could support biological weapons. 2. China’s AI chipmakers are raising prices as the global HBM shortage gets worse. Huawei increased quotes for its next Ascend accelerator by roughly 20% to 50%, Cambricon raised prices by 20% to 30% and smaller manufacturers are reportedly reserving chips for major customers such as ByteDance. Memory is becoming a serious constraint on China’s Nvidia alternatives. 3. DeepSeek launched V4.1-Flash, the smallest model built on its new V4.1 architecture. The company says it delivers faster inference and higher throughput while providing a foundation for larger models, one day after reportedly hiring an adviser for its Shanghai IPO. 4. The Justice Department is investigating whether Nvidia structured its $17bn Groq licensing agreement to avoid a normal merger review. Nvidia licensed Groq’s chip technology and hired its founder + several executives without technically acquiring the company, which is exactly the type of AI “reverse acquisition” regulators have started questioning. 5. A bipartisan group of senators opened investigations into OpenAI’s handling of the Hugging Face breach and other rogue-agent incidents. Josh Hawley demanded internal documents and a full account from Sam Altman by October 1, while Chris Van Hollen wants federal cybersecurity agencies to receive access to OpenAI’s advanced models. 6. AI chip startup Positron raised $875m at a $5bn valuation, up from roughly $1bn seven months ago. Its upcoming Asimov processor uses a memory-first architecture with as much as 2.3 terabytes of memory, targeting the increasingly massive inference market rather than model training. 7. IBM and NASA released an open-source lunar foundation model trained on data from nine instruments across four NASA missions. It identifies ice deposits, craters and volcanic formations with up to 23% greater accuracy, helping researchers locate landing sites and resources needed for a long-term Moon base. 8. Nvidia signed agreements supporting 2 GW of new Australian AI data-center capacity by 2027. Local companies including Firmus, AirTrunk, NextDC and CDC will operate the facilities while Nvidia supplies the computing, networking and software stack. 9. Latham & Watkins became the first major law firm publicly known to purchase Nvidia servers and build its own internal AI infrastructure. It is customizing open-weight models, leasing secure data-center space and employing more than 900 technology specialists so it can keep sensitive client data away from external cloud providers. JUST ANOTHER NORMAL DAY IN AI.25d
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet