Anthropic accuses DeepSeek as Meta Nvidia partner

Key Takeaways

• Anthropic accuses Chinese firms DeepSeek, Moonshot AI, and MiniMax of "distillation" attacks, using 24,000 fake accounts for 16 million interactions to copy its Claude AI model.
• Meta and NVIDIA formed a multiyear partnership, with Meta deploying millions of NVIDIA Blackwell and Rubin GPUs, CPUs, and networking technology for AI infrastructure.
• Google Cloud focuses on three key AI model capabilities: raw intelligence, response time, and cost-effectiveness, leveraging its vertical integration for platforms like Vertex AI.
• Virginia lawmakers postponed most AI legislation until 2027, citing federal regulation conflicts and concerns over losing federal broadband funding.
• UN Secretary General António Guterres raised concerns about AI's impact on resource distribution and the need for human oversight if machines generate most economic output.
• Hollywood faces a conflict between AI investment (e.g., Runway AI, Luma) and concerns from creative professionals and audiences about job displacement and creativity.
• The Coldwater Area Chamber of Commerce will host an "AI for Small Business" webinar on February 24, 2026, to help entrepreneurs integrate AI tools.
• A webinar on February 25, 2025, will guide manufacturers and distributors on using agentic AI for sales, prospecting, and workflow automation.
• India's BharatGen Technology Foundation and Sampige Semiconductors Private Limited partnered to develop sovereign AI hardware and a unified software stack under 'Make in India'.
• Poet Sasha Stiles is showcasing her AI-assisted work, 'A Living Poem,' at MoMA, using a fine-tuned GPT-2 model as a creative collaborator.

Anthropic Accuses Chinese AI Firms of Stealing Models

AI company Anthropic claims that three Chinese firms, DeepSeek, Moonshot AI, and MiniMax, illegally used its Claude AI model. These companies allegedly created over 16 million interactions using about 24,000 fake accounts to 'distill' Claude's capabilities into their own models. Anthropic stated this practice, known as distillation, helps competitors gain advanced AI features faster and cheaper. The company also warned that improperly copied models might lack safety features, posing national security risks. Anthropic is improving its detection systems and sharing information with other AI labs to combat these attacks.

Chinese AI Labs Accused of Stealing Claude Model

Anthropic has publicly accused three Chinese AI labs DeepSeek, Moonshot AI, and MiniMax of conducting large-scale 'distillation' attacks. These companies allegedly used over 24,000 fake accounts and proxy services to access Anthropic's Claude AI model millions of times. The goal was to copy Claude's advanced capabilities without authorization, saving significant development time and cost. Anthropic identified these attacks using IP addresses, metadata, and industry partner information. The company is enhancing its detection methods and collaborating with others to prevent future unauthorized model copying.

Anthropic Claims Chinese Firms Stole AI Models

Anthropic, the creator of the AI chatbot Claude, alleges that three Chinese companies DeepSeek, Moonshot, and MiniMax improperly used Claude to improve their own AI models. These firms reportedly created over 16 million interactions using approximately 24,000 fake accounts, violating Anthropic's terms of service. Anthropic explained that this 'distillation' technique trains a weaker model using a stronger one's outputs. The company warned that such stolen models may lack safety features, creating national security risks. Anthropic is advocating for export controls on chips to limit such activities.

AI Job Losses Raise Questions About Resource Distribution

As artificial intelligence advances and potentially makes human labor obsolete, a critical question arises: how will resources be distributed? This concern goes beyond technical challenges, focusing on the political and power structures that will emerge. If machines generate most economic output, who will decide how wealth is shared? The debate includes designing systems for redistribution and addressing who controls resources. Experts like UN Secretary General António Guterres emphasize the need for human agency and oversight, warning against decisions being made by a few billionaires.

Virginia Tables Most AI Legislation Until 2027

Most artificial intelligence legislation proposed in Virginia has been postponed until 2027. Lawmakers faced challenges, including potential conflicts with federal regulations and concerns about broadband funding. President Trump's executive order suggests withholding funds from states with overly burdensome AI laws. Del. Cliff Hayes Jr. noted that AI bills must align with federal guidelines and avoid hindering the industry. Many bills were shelved due to fears of losing federal broadband funds, impacting the state's ability to regulate AI effectively.

Coldwater Chamber Offers AI Training for Small Businesses

The Coldwater Area Chamber of Commerce is hosting an online webinar titled 'AI for Small Business' on Tuesday, February 24, 2026. The session aims to help local entrepreneurs understand and use artificial intelligence without needing technical expertise. Attendees will learn practical ways to implement AI tools to improve efficiency and creativity in their daily operations. This workshop is part of a larger trend where chambers of commerce provide digital skills training as AI becomes more common in the workplace.

Meta and NVIDIA Forge Major AI Infrastructure Partnership

Meta and NVIDIA have announced a multiyear partnership to build advanced AI infrastructure. Meta will deploy millions of NVIDIA Blackwell and Rubin GPUs, along with NVIDIA CPUs and networking technology, in its hyperscale data centers. This collaboration aims to optimize data centers for both training and inference of AI models. The partnership includes integrating NVIDIA Spectrum-X Ethernet switches into Meta's systems. Both companies highlight the scale of Meta's AI operations and the importance of this collaboration for future AI development.

Google Cloud AI Focuses on Three Key Model Capabilities

Michael Gerstenhaber, VP at Google Cloud, discussed the three main frontiers for AI model capabilities: raw intelligence, response time, and cost-effectiveness for large-scale deployment. He explained that while some applications need the absolute best intelligence, others prioritize speed, and some require affordable models for massive use. Gerstenhaber highlighted Google's vertical integration, from infrastructure to models, as a key strength for developing AI. He noted that Google Cloud's Vertex AI platform provides customers access to advanced models and agentic capabilities.

Hollywood Grapples with AI's Impact on Jobs and Creativity

The rapid growth of AI in Hollywood is creating a conflict between financial investment and creative concerns. Companies like Runway AI and Luma have raised significant funding, while tech giants like Google and ByteDance release new AI models. However, many writers, directors, and consumers express skepticism about AI's impact on jobs and creativity. Recent surveys show negative reactions to AI-generated ads, particularly among younger audiences. The industry faces challenges in balancing technological advancement with the concerns of professionals and the public.

Manufacturers and Distributors to Learn AI Strategies

A webinar on February 25, 2025, will focus on how manufacturers and distributors can navigate uncertainty using AI. The session, featuring Ryan Delahanty and Supplyco AI's Jannik Wiedenhaupt and Henri Freese, will address challenges like tariffs and overseas competition. It will explore how agentic AI can help find and close deals faster, showcasing real applications in prospecting and workflow automation. The webinar aims to provide practical insights rather than a sales pitch, offering a realistic look at current AI successes in manufacturing sales.

India Advances Sovereign AI Hardware with New Partnership

BharatGen Technology Foundation and Sampige Semiconductors Private Limited have signed an agreement to develop India's own AI hardware. This collaboration, led by semiconductor entrepreneur Parag Naik, aims to create AI chipsets and hardware-aware models specifically for India. The initiative falls under the 'Make in India' framework, focusing on building a sovereign AI ecosystem. The partnership will also develop a unified software stack to support these new hardware developments.

Poet Sasha Stiles' AI Work Featured at MoMA

Poet Sasha Stiles is showcasing her AI-assisted work, 'A Living Poem,' at the Museum of Modern Art (MoMA). Stiles uses AI models like GPT-2 as a collaborator, feeding them her own poetry to generate new verses. She fine-tuned a GPT-2 model with her own 200-page manuscript, creating a system that understands her style and themes. 'A Living Poem' is an evolving performance where code, datasets, and human input create a dynamic poetry experience, exploring language as a living system.

Sources

• Anthropic Says DeepSeek, MiniMax Distilled AI Models for Gains
• Anthropic Says That DeepSeek, Moonshot And Minimax Created 24,000 Fake Claude Accounts To ‘Steal’ Its Models
• Chinese companies used Claude to improve own models, Anthropic says
• If AI makes human labor obsolete, who decides who gets to eat?
• Most artificial intelligence legislation in Virginia was tabled until 2027
• Business Beat : Coldwater Chamber to AI training session Tuesday
• Meta builds AI infrastructure with NVIDIA - Electronic Products & Technology
• Google's Cloud AI lead on the three frontiers of model capability
• The Fight Over AI in Hollywood Is a Battle Between Money and Activism
• Navigate Uncertainty with AI - Manufacturing Webinar Feb 26
• BharatGen ties up with Parag Naik's Semiconductor Venture for Sovereign AI Hardware
• How a poet uses AI to write and why her work is now at MoMA