Anthropic unveils security warnings as Nvidia ships H200 chips

Key Takeaways

• Anthropic CEO Dario Amodei criticized the US decision to allow Nvidia to sell H200 AI chips and AMD to sell chips to China, citing national security risks and comparing it to selling nuclear weapons.
• Amodei believes the US holds a multi-year lead in chip manufacturing, and while DeepSeek's R1 model is impressive, China tends to catch up rather than innovate beyond the frontier.
• Over half (56%) of CEOs reported no significant financial benefits from AI investments in the past year, according to PwC's 29th Global CEO Survey.
• Companies with strong "AI foundations," including Responsible AI frameworks, are three times more likely to see positive returns from AI investments.
• The US FDA and European EMA established 10 principles for AI use in drug development, focusing on patient safety, effectiveness, data privacy, and risk assessment.
• Many companies lack proper cybersecurity rules for AI tools, with only 22% of CISOs specifically vetting AI vendors despite 60% identifying new risks.
• OpenAI targets 2026 for widespread "practical adoption" of AI in healthcare, scientific research, and business, planning to launch its own hardware.
• Oklahoma schools, including Broken Arrow High School, are implementing AI courses to prepare students for future jobs, teaching about AI tools, ethics, and bias.
• Parents are advised to monitor children's AI use due to the increasing difficulty in distinguishing real from AI-generated content and the potential for "bad actors" to use AI.
• Businesses often underestimate the preparatory work needed for data and systems to be ready for agentic AI, though tools like Copilot are shown to amplify human intelligence and speed up results.

Anthropic CEO warns against selling AI chips to China

Anthropic CEO Dario Amodei strongly criticized the US decision to allow Nvidia to sell advanced AI chips to China. Speaking at the World Economic Forum in Davos, Amodei compared this move to selling nuclear weapons to North Korea. He believes the US is many years ahead in chip manufacturing and shipping these chips is a big mistake with national security implications. Nvidia, a key partner and investor for Anthropic, stated that offering H200 chips to vetted commercial customers balances competition and American jobs. Amodei also mentioned being impressed by DeepSeek's R1 model but noted China's ability to catch up rather than innovate beyond the frontier.

AI CEO warns on China chip sales

The chief executive of a leading AI company warned that allowing the sale of advanced computer chips to China is a serious risk. He compared this action to selling nuclear weapons to North Korea. This statement highlights deep concerns about national security and technology transfer.

US allows AI chip sales to China despite security warnings

The US administration recently allowed Nvidia to sell its H200 chips and AMD to sell its chips to Chinese customers. This decision has caused controversy due to the chips' use in high-performance AI. Anthropic CEO Dario Amodei strongly criticized the move at the World Economic Forum in Davos. He compared selling these chips to giving nuclear weapons to North Korea, warning of huge national security risks. Amodei emphasized that the US is years ahead in chip manufacturing. Nvidia, a major partner and investor in Anthropic, supplies the GPUs that power Anthropic's AI models.

Anthropic CEO criticizes US chip sales to China

Anthropic's CEO, Dario Amodei, strongly criticized the US decision to allow advanced AI chip sales to China at the World Economic Forum in Davos. He warned that this decision would harm the US, comparing it to selling nuclear weapons to North Korea. Amodei highlighted the significant national security risks of AI, imagining a "country of geniuses in a data center" controlled by one nation. Despite Nvidia being a major partner and investor in Anthropic, supplying GPUs and investing up to $10 billion, Amodei spoke fearlessly. He believes the US is many years ahead in chip manufacturing and shipping these chips is a big mistake.

Businesses invest in AI but struggle to see returns

Many businesses are investing in AI, but most are not yet seeing clear financial benefits, according to PwC's 29th Global CEO Survey. Only about one-third of companies reported real revenue increases or cost savings from AI in the last year. CEOs are eager but cautious, with over half not seeing any financial gains from their AI efforts. The survey found that only a small group of "vanguard" companies are successfully using AI to boost revenue and cut costs. Companies also face challenges with trust, data privacy, and cybersecurity risks related to AI. PwC suggests that businesses need better strategies and partnerships to make AI investments truly pay off.

CEOs frustrated by low AI investment returns

Many CEOs are unhappy with the low financial returns from their AI investments, according to PwC's 29th Global CEO Survey. Over half, 56%, reported no major financial benefits from AI so far. Despite widespread investment in AI, cloud, and data analytics, only 30% of CEOs are confident about revenue growth in 2026. However, companies with strong "AI foundations," like Responsible AI frameworks, are three times more likely to see good returns. Research from MIT also showed that 95% of organizations saw no return on generative AI projects last year. PwC emphasizes that scaling AI beyond pilot projects is crucial for growth and competitive advantage.

FDA and EMA set rules for AI in drug making

The US FDA and European EMA have created 10 principles for using AI in drug development. These guidelines aim to ensure patient safety and drug effectiveness as AI becomes more common. AI can speed up drug development, improve drug monitoring, and reduce animal testing. However, without proper oversight, AI could produce wrong results. The principles cover areas like human-centric design, risk assessment, data privacy, and clear communication. These rules will help guide the safe and responsible use of AI in creating new medicines.

New cyber rules needed for AI tools

Companies are quickly adopting AI tools, but many lack proper cybersecurity rules for them. Only 22% of CISOs have special ways to check AI vendors, even though 60% see new risks. Unlike regular software, AI tools can expose data more widely and use prompt data for training, making it hard to control once entered. Issues like "hallucinations" and "confident lies" from AI also raise trust concerns. Current security processes are too slow for the fast pace of AI adoption, forcing companies to choose between speed and safety. New cyber governance frameworks are needed to manage AI risks effectively and protect against growing third-party cyber incidents.

Boost security with AI for identity and network access

In 2026, companies must improve their security using AI to fight growing cyber threats. Threat actors use AI to automate attacks like phishing and impersonation, making defenses harder. Four key priorities include using AI for fast and adaptive protection, managing and protecting AI agents, extending Zero Trust principles, and strengthening identity security. AI agents can help security teams proactively design and refine access policies, working alongside humans to improve coverage and respond to risks quickly. It is also crucial to treat every AI agent as a unique identity, giving it clear ownership and security standards to prevent "agent sprawl" and data leaks.

Parents should watch kids' AI use

Carla Garrison, technology director for Marshall County Schools, advises parents to monitor their children's use of electronic devices and AI. She warns that with AI, it is becoming very difficult to tell what is real and what is not. Garrison also noted that "bad actors" now have AI tools at their disposal, posing risks to children. Limiting screen time can help overall. The school district is also upgrading its own technology, adding protections for personal information, using Arctic Wolf for virus detection, and updating computers to Windows 11.

Product managers must innovate with AI

The world of product development is changing rapidly because of AI. Companies are using AI to create new products faster and keep customers more engaged. To stay competitive, product managers must focus on smart product changes and making things personal for users. Programs like IIM Kozhikode's Professional Certificate Programme in AI Product Development & Innovation are helping leaders learn these important skills for an AI-focused future.

OpenAI plans practical AI use and new hardware in 2026

OpenAI has marked 2026 as the year for widespread "practical adoption" of AI. The company plans to speed up AI use in important areas like healthcare, scientific research, and business. OpenAI is also looking to launch its own hardware and create new ways to earn money. This strategic push aims to make AI a more common and useful tool across various industries.

Oklahoma schools teach students about AI

Oklahoma schools are now teaching students about artificial intelligence to prepare them for future jobs. Broken Arrow High School started the state's first full AI course in 2026 for juniors and seniors, teaching about AI tools, ethics, and bias. About 150 students are enrolled, with plans to expand the program. The University of Oklahoma Polytechnic Institute also offers the state's first bachelor's degree in applied AI, teaching students what AI can and cannot do. Educators believe these programs are crucial for students to understand and use AI responsibly in a rapidly changing world.

Podcast shares real AI lessons for businesses

The AI Agent & Copilot Podcast shared real lessons about using AI in businesses with Crystal Ahrens. Many organizations often underestimate the work needed to get data and systems ready for agentic AI. The Summit's master classes revealed practical details like costs and staffing models, showing how AI helps people work better rather than replacing them. The discussions highlighted that AI amplifies human intelligence and speeds up results, with tools like Copilot making people more effective.

Sources

• AI boss says chip sales to China ‘like selling nuclear weapons to North Korea’
• AI boss says chip sales to China ‘like selling nuclear weapons to North Korea’
• US Allows Nvidia and AMD AI Chip Sales to China Amid Security Concerns | Ukraine news
• Anthropic's CEO stuns Davos with Nvidia criticism
• Enterprise AI investments are forging ahead despite elusive ROI
• CEOs are fed up with poor returns on investment from AI: Enterprises are struggling to even 'move beyond pilots' and 56% say the technology has delivered zero cost or revenue improvements
• Regulatory Agencies Establish Principles of Good AI Use in Drug Development
• Why AI adoption requires a dedicated approach to cyber governance
• Four priorities for AI-powered identity and network access security in 2026
• Parents Urged To Monitor Kids’ Use of Artificial Intelligence
• Why product managers must rethink innovation in the age of AI
• OpenAI sets 2026 as year for practical AI adoption, eyes hardware debut and new revenue streams
• Oklahoma schools bring artificial intelligence into the classroom
• AI Agent & Copilot Podcast: Real Enterprise AI Lessons with Crystal Ahrens