US AI company Anthropic recently accused three Chinese AI firms—DeepSeek, Moonshot AI, and MiniMax—of illegally copying its Claude model. Anthropic alleges these companies used over 16 million queries generated by approximately 24,000 fake accounts to train their own models through a process known as distillation. This practice, according to Anthropic, violates their terms of service, poses national security risks due to potentially missing safety features, and circumvents export controls. OpenAI has also made similar accusations against DeepSeek, highlighting a growing concern over intellectual property in the AI sector.
The accusations, however, have drawn some criticism, with detractors pointing to a perceived hypocrisy given that AI companies often train models on vast, sometimes unlicensed, datasets. While the Chinese labs reportedly paid for API access, Anthropic maintains that using distillation to gain a competitive advantage from a competitor's model is illicit. In response, Anthropic is enhancing its security measures to detect and prevent such activities and is collaborating with industry partners and authorities to address these sophisticated attacks.
Beyond these intellectual property disputes, AI continues to drive significant shifts across various industries. For instance, AI chatbots are projected to substantially boost eCommerce sales by 2026, enhancing personalization and guiding customers to faster purchase decisions. SAP CX CMO Jessica Keehn suggests that by 2026, AI assistants could largely replace traditional search engines like Google for retail discovery, offering more personalized results based on user intent and behavior.
In terms of infrastructure and development, Composio has open-sourced its Agent Orchestrator framework, designed to help AI developers build scalable multi-agent workflows by separating planning from execution. Harness also launched its Artifact Registry, which integrates artifact management into CI/CD environments, reducing diagnosis time for software artifact issues by 83% and incorporating AI security controls like a Dependency Firewall. Meanwhile, Meta is diversifying its AI infrastructure by deploying AMD's Helios GPUs alongside Nvidia's chips, aiming to reduce dependency on a single supplier and validate AMD's AI accelerator efforts.
Globally, nations are also advancing their AI strategies. South Korea is investing 700 billion won in 'Manufacturing AX' to boost industrial efficiency by 20%, already seeing a 12.5% reduction in welding time through AI integration. Conversely, India's push for sovereign AI, exemplified by Sarvam's models, faces challenges in competing with global leaders, as its models currently rely on foreign foundations and lack open-sourced weights, leading developers to often prioritize globally competitive options.
The impact of AI on the workforce remains a significant topic. JPMorgan Chase CEO Jamie Dimon noted that AI is already reshaping the bank's workforce, leading to "huge redeployment plans" for employees, even as the overall headcount remains stable. Experts warn that rapid AI advancements could lead to widespread job disruption, particularly for white-collar workers, necessitating societal planning and support for displaced individuals to prevent potential political instability. Tools like Kreado AI, which offers multilingual video creation with over 700 virtual avatars and 1,400 languages, demonstrate how AI is also transforming content production, making scalable video accessible for marketing and training.
Key Takeaways
- Anthropic accuses Chinese firms DeepSeek, Moonshot AI, and MiniMax of stealing Claude model capabilities using over 16 million queries from 24,000 fake accounts via 'distillation'.
- Anthropic claims this data theft violates terms of service, poses national security risks, and circumvents export controls, with OpenAI making similar accusations against DeepSeek.
- AI chatbots are projected to significantly boost eCommerce sales by 2026 through personalization and guided selling, with AI assistants potentially replacing search engines like Google for retail discovery.
- Composio open-sourced its Agent Orchestrator framework for scalable multi-agent AI workflows, separating planning and execution.
- Harness launched Artifact Registry, integrating artifact management into CI/CD, reducing diagnosis time by 83% and including AI security controls like Dependency Firewall.
- Meta is diversifying its AI infrastructure by deploying AMD's Helios GPUs alongside Nvidia's chips to avoid single-supplier dependency.
- South Korea is investing 700 billion won in 'Manufacturing AX' to achieve a 20% efficiency boost and a 12.5% reduction in welding time through AI.
- India's sovereign AI efforts, like Sarvam's models, face challenges competing globally due to reliance on foreign foundations and lack of open-sourced weights.
- JPMorgan Chase CEO Jamie Dimon states AI is leading to "huge redeployment plans" for employees, while experts warn of widespread job disruption, especially for white-collar workers.
- Kreado AI offers a platform for creating multilingual videos with over 700 virtual avatars and 1,400 languages, making scalable video production accessible.
Anthropic accuses Chinese AI firms of massive data theft
AI company Anthropic claims three Chinese firms, DeepSeek, Moonshot AI, and MiniMax, illegally copied its Claude model. They allegedly used over 16 million queries from 24,000 fake accounts to train their own smaller AI models. Anthropic stated this practice violates their terms of service and could pose national security risks. The company is improving its defenses and sharing information with other AI labs and authorities to prevent future theft.
US AI firm Anthropic alleges Chinese rivals stole data
US AI company Anthropic accuses three Chinese AI labs, DeepSeek, Minimax, and Moonshot AI, of illegally copying its Claude model. They allegedly used over 16 million interactions from 24,000 fake accounts to train their own models, a process called distillation. Anthropic warns this could create national security risks if these models lack safety features. This follows similar accusations by OpenAI against DeepSeek.
Chinese AI firms allegedly used 16 million Claude queries for theft
Anthropic reported that three Chinese AI companies, DeepSeek, Moonshot AI, and MiniMax, illegally copied its Claude AI model. They allegedly used over 16 million queries from about 24,000 fake accounts to train their own models through a technique called distillation. Anthropic stated these actions violate their terms of service and could lead to national security risks due to potentially missing safety features. The company is enhancing its security measures to detect and prevent such activities.
Anthropic details AI data theft by Chinese firms
Anthropic has identified large-scale campaigns by Chinese AI labs DeepSeek, Moonshot, and MiniMax to steal capabilities from its Claude model. These firms allegedly used over 16 million interactions from roughly 24,000 fake accounts to train their own models via distillation. Anthropic believes this illicitly distilled data poses national security risks because it may lack safety features. The company is working with industry partners and authorities to combat these attacks.
Anthropic accuses Chinese companies of stealing AI data
AI company Anthropic has accused Chinese firms of illegally copying data from its Claude chatbot to improve their own AI models. This practice, known as distillation, involves training a smaller model using outputs from a larger one. Anthropic stated that this unauthorized data extraction raises concerns about intellectual property theft and national security. Similar accusations were previously made by OpenAI against Chinese AI firms.
US AI giant Anthropic claims Chinese rivals stole data
US AI company Anthropic alleges that three Chinese firms, DeepSeek, Moonshot AI, and MiniMax, have stolen capabilities from its Claude chatbot. They used a technique called 'distillation,' involving about 16 million interactions from 24,000 fake accounts, to boost their own AI models. Anthropic warns this theft circumvents export controls and poses national security risks due to potentially missing safety features. OpenAI has made similar accusations against Chinese companies.
Anthropic reveals Chinese AI firms' data theft methods
Anthropic has detailed how three Chinese AI labs, DeepSeek, Moonshot, and MiniMax, allegedly used 'distillation' to steal capabilities from its Claude model. They reportedly created over 24,000 fake accounts to generate 16 million interactions, violating terms of service. Anthropic emphasizes that distillation can be legitimate but is illegal when used by competitors to gain an advantage. The company is calling for industry-wide action to address these sophisticated attacks.
Anthropic, OpenAI flag Chinese AI firms' data theft
Anthropic accused three Chinese AI firms, DeepSeek, Moonshot AI, and MiniMax, of using 'distillation attacks' to copy its Claude model. These firms allegedly used commercial proxy services and tens of thousands of fake accounts to generate over 16 million interactions. Anthropic and OpenAI are concerned that this practice allows Chinese firms to gain a competitive advantage and poses national security risks. Distillation is a common AI training method, but using it on competitor models is banned.
Backlash over Anthropic's claims on Chinese AI data theft
Anthropic's accusation that Chinese AI labs DeepSeek, Moonshot AI, and MiniMax stole data via distillation has drawn criticism. Critics accuse Anthropic of hypocrisy, noting that AI companies train models on vast, often unlicensed, data. While the Chinese labs paid for API access, critics argue this is more legitimate than Anthropic's alleged use of unlicensed data. Anthropic defends its position by citing contractual violations and potential national security risks from unprotected AI models.
AI chatbots boost eCommerce sales significantly in 2026
AI chatbots are significantly increasing eCommerce sales in 2026 by enhancing personalization and guided selling. These tools help customers make faster, more confident purchase decisions, leading to higher conversion rates and order values. Recovering abandoned carts is a key function, with chatbots re-engaging users at crucial moments. Successful implementation requires strong integration with product data and CRM systems, with results measured by business KPIs like conversion uplift and revenue recovery.
AI assistants may replace search engines by 2026
AI assistants are poised to replace traditional search engines for retail discovery, according to SAP CX CMO Jessica Keehn. By 2026, most shoppers may use AI agents instead of search engines like Google. AI-powered search understands user intent and context, offering personalized results based on behavior and preferences. This shift moves product discovery from keyword matching to AI recommendations, providing a more tailored shopping experience.
Composio releases Agent Orchestrator for AI developers
Composio has open-sourced its Agent Orchestrator framework to help AI developers build scalable multi-agent workflows. This tool separates planning from execution, allowing a 'Planner' to break down tasks and an 'Executor' to handle tool interactions. It uses 'Managed Toolsets' to provide agents with only necessary information, reducing confusion and improving accuracy. The framework also offers state management for better error handling and traceability in AI agent systems.
Harness launches Artifact Registry with AI security
Harness has released Artifact Registry, a platform that integrates artifact management into the CI/CD environment. This reduces the time to diagnose and fix software artifact issues by 83 percent. The registry supports various formats like Docker, Python, and npm, and includes AI security controls. Its Dependency Firewall blocks risky components at intake, and artifact quarantine isolates items that fail security checks, enhancing software supply chain security.
India's sovereign AI efforts face challenges
India's push for sovereign AI, exemplified by Sarvam's models, faces challenges in competing with global AI leaders. While Sarvam aims to serve India's linguistic diversity, its models currently rely on foreign foundations and lack open-sourced weights. Developers often prioritize globally competitive models for scalability and capability over national origin. This pattern mirrors past Indian tech products that struggled due to prioritizing national pride over market competitiveness.
South Korea invests 700 billion won in AI for manufacturing
South Korea is investing 700 billion won in 'Manufacturing AX' to boost efficiency by 20% through AI. This initiative is already reducing welding time by 12.5% and involves process optimization, AI quality control, predictive maintenance, and integrating robots, including humanoid ones. The program aims to enhance manufacturing competitiveness and expand AI use into service sectors.
JPMorgan CEO: AI is reshaping workforce, redeploying staff
JPMorgan Chase CEO Jamie Dimon stated that AI is already changing the bank's workforce, leading to 'huge redeployment plans' for employees. While the bank's overall headcount remains stable, AI is displacing some roles, and the bank is offering affected employees other jobs. JPMorgan is significantly increasing its use of generative AI, focusing on customer service and technology roles, and is committed to deploying AI to improve customer service.
AI advancements threaten jobs, experts warn
Rapid AI advancements are surpassing human intelligence in many tasks, potentially leading to widespread job disruption, particularly for white-collar workers. Experts warn this could destabilize politics if millions become unemployable. However, there's a call to establish rules governing AI use to prevent a dystopian future. The impact of AI on jobs is a growing concern that requires societal planning and support for displaced workers.
Kreado AI offers multilingual videos with virtual avatars
Kreado AI provides a platform for creating multilingual videos using virtual avatars, addressing the high cost and complexity of traditional video production. It offers over 700 avatars and supports more than 1,400 languages and voices, including AI voice cloning. The tool can convert text or URLs into video drafts with virtual presenters. Kreado AI aims to make scalable, multilingual video production accessible for marketing and training.
Meta uses AMD chips alongside Nvidia for AI
Meta is diversifying its AI infrastructure by using AMD's Helios GPUs alongside Nvidia's chips. This multi-vendor strategy aims to avoid dependency on a single supplier. The deal validates AMD's AI accelerator efforts and puts pressure on Nvidia's market dominance. Meta's large-scale deployment of both chip types signals a maturing AI infrastructure market demanding choice, performance, and competitive pricing.
Sources
- Anthropic accuses DeepSeek, other Chinese AI developers of 'industrial-scale' copying — Claims 'distillation' included 24,000 fraudulent accounts and 16 million exchanges to train smaller models
- Are China’s ‘AI tigers’ cheating? US rival Anthropic alleges some are
- Anthropic Says Chinese AI Firms Used 16 Million Claude Queries to Copy Model
- Detecting and preventing distillation attacks
- Anthropic Accuses Chinese Companies of Siphoning Data From Claude
- US AI giant accuses Chinese rivals of mass data theft
- Anthropic exposes how Chinese AI firms try to steal LLM tech
- Anthropic joins OpenAI in flagging 'industrial-scale' distillation campaigns by Chinese AI firms
- Hypocrite Or Victim? The Backlash To Anthropic’s Attack On Chinese AI Labs Over Distillation, Explained
- How AI Chatbots for eCommerce are Driving 3x More Sales in 2026
- Traditional Research Is Out: Could This Be The Last Year We Use Search Engines?
- Composio Open Sources Agent Orchestrator to Help AI Developers Build Scalable Multi-Agent Workflows Beyond the Traditional ReAct Loops
- Harness launches Artifact Registry with AI security controls
- India’s Sovereign AI Trap: National Pride Meets Developer Pragmatism
- Manufacturing AX: 700 Billion Won Investment Boosts Efficiency 20%
- Jamie Dimon says AI is already reshaping JPMorgan Chase's workforce as bank plans 'huge redeployment’
- AI: Yes, it’s coming for your job
- Kreado AI Testbericht 2026: Mehrsprachige KI-Videos mit virtuellen Avataren
- Meta Taps AMD Chips Days After Nvidia Deal, Shaking Up AI Race
Comments
Please log in to post a comment.