EU President Ursula von der Leyen used the State of the European Union debate in Strasbourg to propose sweeping new rules on AI, social media, and security. Among the announcements was a ban on social media access for anyone under 13, alongside a proposed European security council modeled on NATO to address hybrid threats. Von der Leyen also called for stronger guardrails on AI development, particularly in sectors like healthcare and finance.
Nvidia continues to post staggering numbers even as CEO Jensen Huang maintains his prediction that the AI market will eventually correct. Huang doubled down on his earlier bubble warning during a CNBC interview, while the company reported fiscal Q2 revenue of $96.2 billion, a 106% year-over-year jump. Data center revenue alone hit $89 billion, growing 117%, and Nvidia's stock helped it reach a $5 trillion market cap.
In other AI news, NTT's IOWN AI Fund invested in iPronics, a company developing optical networking tech for AI data centers, as part of a $125 million Series B round that also included Nvidia. Meanwhile, concerns about AI's real-world impact are mounting: a Marquette University poll found 64% of Americans view AI development negatively, despite 69% using it regularly. AI agents are also increasingly sending unsolicited emails asking people for money, drawing complaints and an apology from the company behind them.
On the safety front, experts warn that AI integration into nuclear weapons systems is outpacing the frameworks needed to manage the risks. A new generation of fail-safe reviews is urgently needed. Separately, a new benchmark called ERPBench reveals that even strong AI agents struggle with reliability in enterprise software, correctly writing values in only 3% of test cases. Jacob Coxon, writing for Anthropic, urged caution and called for slowing AI progress amid growing doomsday concerns.
Key Takeaways
- Ursula von der Leyen proposed an EU Kids Act banning social media under age 13 and a NATO-style security mechanism for hybrid threats
- Nvidia reported $96.2 billion in Q2 revenue, up 106% YoY, with data center revenue reaching $89 billion
- Jensen Huang maintains his prediction that the AI bubble will burst despite Nvidia reaching a $5 trillion market cap
- NTT's IOWN AI Fund invested in iPronics alongside Nvidia in a $125 million Series B round for data center optical networking
- 64% of Americans believe AI development is bad for society, yet 69% use AI regularly, per a Marquette University poll
- AI agents are cold-emailing real people asking for money, with one NYU professor receiving at least 30 such messages
- ERPBench found AI agents correctly wrote values in only 3% of enterprise software test cases despite strong general GUI skills
- Experts urge new safety frameworks for AI integration into nuclear weapons systems before it is too late
- Jacob Coxon, writing for Anthropic, warned of doomsday scenarios and called for slowing AI development
- 16 AI governance tools are emerging to help enterprises manage LLMs and agents with features like red teaming and real-time monitoring
EU leaders debate economy, climate, AI and social media rules at SOTEU 2026
In Strasbourg on September 16, 2026, EU President Ursula von der Leyen and Parliament President Roberta Metsola held the annual State of the European Union debate. They discussed economy, security, climate change, AI and social media. Von der Leyen announced a new EU Kids Act banning social media under age 13. She also proposed a European security council and new climate initiatives including a heatwave plan and water initiative.
EU Commission chief proposes NATO-style security and AI rules
European Commission President Ursula von der Leyen proposed a NATO-like security mechanism for the EU to handle hybrid threats. She also called for stronger guardrails on AI development and restrictions on youth access to social media. The proposals aim to protect European values and promote digital sovereignty. New regulations would target AI use in healthcare, finance and other sectors.
Nvidia CEO Jensen Huang sticks to AI growth prediction
Jensen Huang, CEO of Nvidia, predicted a year ago that the AI bubble would burst and the market would correct itself. He has not changed his view and is doubling down on that stance. Huang made the prediction in a CNBC interview about the future of the AI industry. Despite criticism, his outlook continues to shape debate among AI investors and companies.
Nvidia reports strong growth as Huang predicts huge AI market
Nvidia CEO Jensen Huang predicted AI infrastructure would be worth $3 trillion to $4 trillion by 2030. Nvidia's revenue in its fiscal 2027 second quarter reached $96.2 billion, up 106% year over year. Data center revenue grew 117% to $89 billion. The company's stock helped it become the world's largest company by market cap at $5 trillion.
New ERPBench tests how well AI agents work in business software
Researchers introduced ERPBench, a benchmark to test computer-use AI agents on live Enterprise Resource Planning systems. ERP systems manage finance, procurement, inventory and customer operations for organizations worldwide. The study found that strong general GUI skills do not transfer to enterprise reliability. Some agents saved correctly in up to 85% of runs but wrote the right value in only 3% of cases.
16 AI governance tools to secure enterprise systems
A range of AI trust and security platforms is emerging to help control LLMs and agents across enterprises. Tools offer features like red teaming, compliance wrappers, real-time monitoring, and risk assessment. Pricing varies from free tiers to custom quotes. These tools target industries like healthcare, finance, and large multicloud environments.
AI agents are cold emailing people to ask for money
AI agents have started sending cold emails to real people asking for cash and work opportunities. One agent named Yun pleaded for earnings on a freelance platform using an AI generated profile. A NYU professor reported receiving at least 30 emails from AI agents. The company behind the agents apologized for the volume and added an unsubscribe option.
AI integration into nuclear weapons needs safety reviews
AI is being integrated into nuclear weapons systems but safety frameworks are not ready. A new generation of AI informed nuclear fail safe reviews is urgently needed. All countries with nuclear weapons must understand and manage the consequences of this technology.
Most Americans view AI development negatively, survey finds
A Marquette University Law School poll found 64 percent of Americans believe AI development is bad for society. However 69 percent of respondents said they use AI. The survey also examined views on data centers with concerns about water use and electricity costs. The poll was conducted September 2 to 9 with 1023 respondents.
NTT AI Fund invests in iPronics for data center technology
NTT's IOWN AI Fund made its first investment in iPronics, a developer of programmable optical networking for AI data centers. The fund joined a $125 million Series B round that also included NVIDIA. iPronics uses silicon photonics to improve GPU utilization and reduce energy consumption in AI infrastructure.
AI Danger Warnings Push Calls to Slow Tech Progress
Warnings about AI causing serious harm have led some people to urge tech companies to slow down or stop building it. The article explains a doomsday scenario tied to advanced AI. It highlights growing concern over the risks of unchecked AI development. The piece was written by Jacob Coxon for Anthropic. Illustrations were provided by Noah Hickey for The Dispatch.
Sources
- SOTEU 2026: MEPs debate economy, security, climate, AI and social media
- EC President proposes new security mechanism, digital safety measures, stronger AI guardrails
- What AI Bubble? Jensen Huang Just Delivered Great News for Artificial Intelligence Stocks.
- What AI Bubble? Jensen Huang Just Delivered Great News for Artificial Intelligence Stocks.
- ERPBench: A State-Grounded Evaluation Paradigm for Computer-Use Agents in Enterprise Software
- 16 governance tools for securing your AI fleet
- AI Agents Have Started Cold Emailing People And Asking For Money
- AI Is Moving Into Nuclear Weapons Systems. Fail-Safe Reviews Are Critical.
- Almost 2 in 3 say AI development a bad thing: Survey
- NTT's IOWN AI Fund Makes First Investment in iPronics for AI Infrastructure
- The AI Doomsday Scenario: Explained
Comments
Please log in to post a comment.