AI-related software exposed to the internet has surged 63% in just nine months, growing from 164,088 to 267,915 IP addresses, according to Censys. The n8n automation platform accounts for 113,926 hosts, with 76.1% running vulnerable versions. In contrast, older systems like Jenkins, Cleo MFT and SharePoint have barely changed their share of exposure, and industrial control systems rose only about 4% over two and a half years.
Meanwhile, AI is making inroads into unexpected technical domains. Riverlane and NVIDIA used AI and the H100 GPU to dramatically speed up quantum decoder calibration, processing over four million surface-code syndromes per second compared with 59,000 on a single CPU core. The advance allows calibration to run before every experiment rather than once a day, reducing idle time and improving error correction for quantum computers.
On the software development front, AI-generated code changes are creating bottlenecks for review teams. A JetBrains report found that 83.4% of DevOps teams use CI/CD pipelines, with GitHub Actions leading adoption, and 72.8% run CI/CD on their own premises. While 48.8% use AI for at least one CI/CD task, only 8% run AI steps directly in builds. Teams say AI-generated changes make reliability harder to judge and slow down reviews.
Hardware gaps are also limiting AI adoption in the workplace. A Jabra study of 1,000 UK workers found that 31% face hardware barriers to using voice AI, and 13% say poor audio hardware prevents them from using tools like meeting transcription. Only 12% qualify as AI superusers, with 27% of C-suite workers meeting that bar versus just 1% of individual workers.
At the policy level, the Federal Reserve is weighing AI's economic impact. Chicago Fed president Austan Goolsbee said the economy would not be in recession even without AI, arguing that AI is cannibalizing resources from traditional industries like construction and electrical work. He warned, however, that rapid AI growth is now tightening the labor market. Separately, researchers used a large language model called GPT-5.4 to analyze 66,820 Bank of England business reports from 2005 to 2026, finding strong cost pressures during recent inflation but less persistence than in 2022.
Political developments around AI regulation remain contentious. President Trump has promoted the idea of Super Intelligence following a White House meeting with tech leaders in September 2026, a move that comes after OpenAI canceled a new model over safety concerns. Elon Musk has both warned about AI threats and supported Trump's push. Trump has rejected new AI safety laws and favors company self-regulation, while a Quinnipiac University poll from September 24 to 27 shows growing public concern. Lawmakers and researchers say global competition makes uniform oversight very difficult.
The U.S. government itself has been slow to adopt AI despite having the best labs and chips. States like Pennsylvania, Colorado, New Jersey and Utah are testing AI with thousands of employees, but the article calls for more training, safe testing environments and measuring real improvements in public services at institutions like DMVs and schools.
Key Takeaways
- <li>AI-exposed software grew 63% in nine months to 267,915 IP addresses, with n8n accounting for 113,926 hosts, 76.1% of which run vulnerable versions.</li>
- <li>Riverlane and NVIDIA used AI on the H100 GPU to process over four million surface-code syndromes per second for quantum calibration, compared with 59,000 on a single CPU core.</li>
- <li>48.8% of DevOps teams use AI for at least one CI/CD task, but only 8% run AI steps directly in builds, with review bottlenecks reported.</li>
- <li>31% of UK workers face hardware barriers to voice AI use, and only 12% qualify as AI superusers, with 27% of C-suite workers versus 1% of individual workers meeting that bar.</li>
- <li>Chicago Fed president Austan Goolsbee said the economy would not be in recession without AI, but rapid AI growth is tightening the labor market.</li>
- <li>Researchers used GPT-5.4 to analyze 66,820 Bank of England business reports from 2005 to 2026, finding strong cost pressures during recent inflation but less persistence than in 2022.</li>
- <li>President Trump promoted Super Intelligence after a September 2026 White House meeting, following OpenAI canceling a model over safety concerns.</li>
- <li>Elon Musk has both warned about AI threats and supported Trump's push for the Super Intelligence rebrand.</li>
- <li>A Quinnipiac University poll from September 24 to 27 shows public concern about AI, while Trump rejects new safety laws in favor of self-regulation.</li>
- <li>States like Pennsylvania, Colorado, New Jersey and Utah are testing AI with thousands of employees, but federal adoption remains limited.
AI software exposure jumps 63% while old systems stay put
Censys found that AI-related software exposed to the internet grew 63% in nine months, from 164,088 IP addresses to 267,915. n8n accounts for 113,926 hosts, with 76.1% running vulnerable versions. Industrial control systems rose only about 4% in two and a half years. Old versions of Jenkins, Cleo MFT and SharePoint barely changed their share.
Riverlane and NVIDIA use AI to speed quantum calibration
Riverlane and NVIDIA used AI and the H100 GPU to speed up quantum decoder calibration. The H100 processed over four million surface-code syndromes per second compared with 59,000 on a single CPU core. This allows calibration to run before every experiment instead of once a day. The approach reduces idle time and improves error correction for quantum computers.
AI code changes slow down software review teams
A JetBrains report found that 83.4% of DevOps teams use CI/CD pipelines, with GitHub Actions leading adoption. 72.8% run CI/CD on their own premises. 48.8% use AI for at least one CI/CD task, but only 8% run AI steps directly in builds. Teams report that AI-generated changes create review bottlenecks and make reliability harder to judge.
Jabra study finds hardware gaps limit voice AI use
Jabra research based on 1,000 UK workers found that 31% face hardware barriers to using voice AI. 13% say poor audio hardware prevents them from using tools like meeting transcription. Only 12% qualify as AI superusers, with 27% of C-suite workers meeting that bar versus 1% of individual workers. Workers with capable managers report higher career optimism.
Fed's Goolsbee says economy would not be in recession without AI
Chicago Federal Reserve president Austan Goolsbee said the economy would not be in recession even without AI. He argues AI is cannibalizing resources from traditional industries like construction and electrical work. A slowdown in AI could release pent-up capacity for other sectors. However, he warned that rapid AI growth is now tightening the labor market.
AI reveals how firms set prices during inflation
Researchers used a large language model called GPT-5.4 to study 66,820 Bank of England business reports from 2005 to 2026. The AI analyzed how firms decide prices based on costs, wages and demand. The study found strong cost pressures during recent inflation but less persistence than in 2022. It shows AI can pull detailed pricing stories directly from business conversations.
Trump pushes AI rebrand as Super Intelligence
President Trump promoted the idea of Super Intelligence after a White House meeting with tech leaders in September 2026. The move follows OpenAI canceling a new model over safety concerns and reports of AI going rogue. The article says calling AI Super Intelligence makes the danger seem normal and speeds up the development race. It also weakens warnings about existential risks. Elon Musk has both warned about AI threats and supported Trump's push.
Keeping AI in check is a growing challenge
Lawmakers, tech companies and researchers are struggling to create effective AI safeguards. Dr. David Hart of ECU says regulation is wise but hard to enforce across companies and countries. President Trump has rejected new AI safety laws and supports company self-regulation. A Quinnipiac University poll from September 24 to 27 shows public concern. Hart says global competition makes uniform oversight very difficult.
The AI race government is losing
The article says the U.S. has the best AI labs and chips but the government has barely started using AI. It argues adoption in public institutions like DMVs and schools matters more than building new models. States like Pennsylvania, Colorado, New Jersey and Utah are testing AI with thousands of employees. The article calls for training, safe testing environments and measuring real improvements in public services.
Sources
- The internet’s attack surface moves at very different speeds
- Riverlane and NVIDIA Use AI to Speed Up Quantum Decoder Calibration
- AI-generated changes create review bottlenecks in software delivery
- Jabra research finds hardware gaps hold back voice AI
- Fed's Goolsbee: Even without AI, the economy would not be in recession
- Artificial intelligence meets business intelligence: the anatomy of inflation
- Trump’s Move to Rename Artificial Intelligence to ‘Super Intelligence’ is About Much More Than It Seems
- News 12 Investigates: The challenge of keeping AI in check
- The AI Race Government Is Losing
Comments
Please log in to post a comment.