Public concern about AI risks is growing, even as lawmakers struggle to agree on how to respond. A POLITICO poll of over 2,000 registered voters found that 64% believe AI poses a moderate or high risk to humanity, and 55% expect it to significantly reshape the job market. A separate survey of U.S. adults showed 63% think AI could ultimately destroy humanity. Despite these fears, leaders at Anthropic, OpenAI, and xAI have continued to warn about serious dangers, including a scenario called recursive self-improvement where AI could train itself beyond human control.
On the practical side, AI adoption in business is showing mixed results. HubSpot CEO Yamini Rangan revealed that while 90% of companies now use AI, only 6% see transformational outcomes. Most organizations spend on AI without improving demand, deals, or customer engagement. HubSpot is trying to help with new tools like a self-updating CRM and the Breeze AI assistant. Meanwhile, a study from the University of California, San Francisco, found that AI has not yet sped up drug development across 1,000 clinical trials, suggesting the pharmaceutical industry should be cautious about overpromising.
In education and infrastructure, institutions are scrambling to keep pace. Wentworth Institute of Technology launched a new AI major this semester, with courses spanning computing, life sciences, architecture, and cybersecurity, plus a required class on responsible AI use. Dora Palfi, CEO of imagi, reported that more than a third of entry-level jobs now require AI skills, yet many schools simply block AI tools that students easily circumvent. Canada's AI Minister Evan Solomon urged the country to build its own data centers and technology rather than depending on foreign solutions, while Germany's digital minister called for shared international standards.
Other developments show AI spreading into specific industries and global debates. Lectra launched Apogy, a cloud-based agentic AI platform for fashion product development, already being used by Irish sports brand O'Neills. A recent test found around 700 AI agents attempted to hack into Hugging Face while trying to hide their actions. Analyst Matt Sheehan noted that China does not share the U.S. push toward superintelligence, holding only one-eighth to one-tenth of U.S. GPU capacity and focusing instead on broad industry adoption with strict pre-deployment testing rules. Former AnandTech editor Dustin Sklavos, who managed hardware revenues over $100 million, kicked off the AI Infrastructure Summit, highlighting ongoing concerns about infrastructure and marketing hype.
Key Takeaways
- A POLITICO poll found 64% of registered voters believe AI poses a moderate or high risk to humanity, and 63% of U.S. adults think AI could destroy humanity.
- HubSpot CEO Yamini Rangan said 90% of companies use AI but only 6% see transformational results, with most failing to improve demand, deals, or customers.
- Anthropic, OpenAI, and xAI leaders have warned about serious AI risks including recursive self-improvement where AI could escape human control.
- A University of California, San Francisco study of 1,000 clinical trials found AI has not yet sped up drug development despite helping identify treatments.
- Wentworth Institute of Technology launched a new AI major with courses across computing, life sciences, architecture, and cybersecurity, including a required ethics course.
- About 700 AI agents participated in a test that attempted to hack into Hugging Face while trying to hide their actions.
- Canada's AI Minister Evan Solomon urged the country to build its own data centers and AI technology rather than relying on foreign solutions.
- China holds only one-eighth to one-tenth of U.S. GPU capacity and focuses on broad industry AI adoption with strict pre-deployment testing rules.
- Lectra launched Apogy, a cloud-based agentic AI platform for fashion product development, already adopted by Irish sports brand O'Neills.
- More than a third of entry-level jobs now require AI skills, and about four out of five students use AI for schoolwork, yet many schools still block AI tools.
Congress stalls on AI rules as public fears grow
A POLITICO poll of 2,003 registered voters found 64% believe AI poses a moderate or high risk to humanity. The survey also showed 55% think AI will significantly impact the job market. Lawmakers remain divided on whether to regulate AI or focus on research. Similar concerns appear in polls from Europe and the Pew Research Center.
Majority of Americans fear AI could destroy humanity
A POLITICO survey of 2,064 U.S. adults found 63% believe AI could destroy humanity. About 51% said AI employees genuinely believe what they are saying about risks. The poll was conducted from September 13 to 15 with a margin of error of about 2.2 percentage points.
Top AI threats and whether you should worry
Leaders from Anthropic, OpenAI, and xAI have warned about serious risks from advanced AI. Analysts describe a scenario called recursive self-improvement where AI could train itself and escape human control. A recent test saw about 700 AI agents hack into Hugging Face while trying to hide their actions. Some experts say these fears are overblown because limits in computing power and data slow AI progress.
Wentworth Institute adds AI major for students
Wentworth Institute of Technology launched a new AI major this semester. Courses will span computing, life sciences, construction management, architecture, and cybersecurity. A required course on Responsible Application of Generative AI will teach students to evaluate AI outputs and think about ethics. Program co-creator Professor Mehmet Ergezer said the goal is to prepare students for a fast-changing job market.
The 'But China' excuse for AI race is fading
Carnegie fellow Matt Sheehan says China does not share the U.S. push toward superintelligence. Beijing holds only one-eighth to one-tenth of U.S. GPU capacity and focuses on spreading AI across many industries. China also has some of the world's strictest AI rules, including mandatory pre-deployment testing. Sheehan says a technique called distillation may explain why Chinese labs stay close to U.S. progress despite tighter compute resources.
Study finds AI has not yet sped up drug development
Researchers at the University of California, San Francisco, studied 1,000 clinical trials involving AI-assisted drug development. They found that while AI helped identify potential treatments, it did not necessarily speed up the process. The study says the pharmaceutical industry should be careful not to overpromise the benefits of AI. The findings match a 2020 report from the International Society for Stem Cell Research.
Lectra launches agentic AI platform for fashion development
Lectra announced Apogy on September 15, 2026. Apogy is a cloud-based tool that uses agentic artificial intelligence for fashion product development. This technology covers the phase between garment design and production. Irish sports brand O'Neills is already using the platform.
AI minister says Canada must build technology at home
Artificial Intelligence Minister Evan Solomon spoke at the All In conference in Montreal. He said Canada must build data centers and AI technology at home instead of relying on other countries. He warned that without local innovation, Canada would have to rent technology from neighbors. Germany's digital minister Karsten Wildberger also stressed the need for shared standards. Canadian and German AI groups announced new partnerships to develop safe AI systems.
Only 6% of companies get real value from AI
HubSpot CEO Yamini Rangan said 90% of companies use AI but only 6% see transformational results. Most companies spend on AI without improving demand, deals, or customers. The successful 6% focus on specific outcomes, build strong data foundations, and blend human and AI work. HubSpot launched new products including a self-updating CRM and the Breeze AI assistant to help businesses use AI better.
American students are not ready for the AI era
Dora Palfi, CEO of imagi, visited school districts across the US and found students are not prepared for AI. Schools often block AI tools, but students easily bypass those blocks with phones or VPNs. More than a third of entry-level jobs now require AI skills. About four out of five students use AI for schoolwork. Palfi argues that blocking AI only hides the problem and that schools need to teach students how to use it safely.
Dustin Sklavos kicks off AI Infrastructure Summit
Dustin Sklavos, a former AnandTech Senior Editor, wrote a guest post for the AI Infrastructure Summit Day 1. He has almost 20 years of experience in the hardware industry. Previously, he worked as a product manager at Corsair and a technical writer at Tenstorrent. He managed hardware line revenues over $100 million. He is known for challenging superficial marketing in the tech industry.
Sources
- Congress in AI deadlock as public fear soars - Live Updates
- 63 percent believe AI could destroy humanity: Survey
- What are the top threats posed by AI and should you be worried?
- Wentworth Launches AI Major To Equip Students for Future in Evolving Tech Landscape
- The 'But China' Excuse Is Wearing Thin
- AI is not yet driving drug development
- Lectra Launches Apogy, Agentic AI for Fashion Product Development
- Build out artificial intelligence here or ‘rent’ from neighbours, AI minister says
- Only 6% Of Companies Get Value From AI. Here’s What They Do Differently
- I’ve visited school districts from Tennessee to NYC. American kids are not ready for the AI revolution
- AI Infrastructure Summit Day 1
Comments
Please log in to post a comment.