Andy Barkett, a former engineering manager at Google and Nvidia, warns against overhyping the risk of AI extinction. He argues that assigning a 10 percent chance to AI killing humanity is misleading and says the technology could bring prosperity or serious disruption. Barkett is more concerned about AI worsening existing problems like misinformation and job loss than about existential threats. He also says government regulation should focus on accountability rather than slowing down progress.
Australia is moving toward a systems-based approach to regulate frontier AI. Assistant minister Andrew Charlton says companies would need to demonstrate rigorous processes for finding and managing AI risks. He argues that voluntary codes fail and detailed rules become outdated quickly. The proposed framework borrows models from workplace safety and banking supervision to keep pace with rapid technological change.
Identity security for AI agents emerged as a top concern at SailPoint's Navigate 2026 event. There are now 109 machine identities for every human, and agents can gain access to systems unintentionally. AWS reported that AgentCore usage grew 15-fold in six months. A Fortune 500 company discovered 10,000 hidden agents it did not know about. The industry focus is shifting toward just-in-time access and stronger human oversight.
Federal AI programs are already affecting real people in unexpected ways. A pilot program uses AI to review and deny Medicare claims for seniors in Washington and five other states. Virtix Health, one of the companies running the program, is paid based on how many claims it denies. In early 2026, the company denied more claims than it approved for 17 procedures, including nerve stimulation and cervical fusion surgery. Doctors and lawmakers say the delays harm patients' health and quality of life.
AI is also reshaping national security and workplace dynamics. A report says AI and data centres are now on the frontline of national defence, with the AI cybersecurity market expected to grow by nearly 25 percent each year until 2033. A 2026 exercise reportedly used Anthropic's Mythos AI model to find government vulnerabilities in hours. Meanwhile, Anthropic CEO Dario Amodei has shared views on how AI will reshape American work, and the shift is already visible across many industries. The TV show Law and Order explored similar dangers in its 26th season premiere, depicting a world where compromised AI systems can manipulate the justice system.
Key Takeaways
- Former Google and Nvidia executive Andy Barkett calls a 10 percent extinction risk for AI misleading and warns against overhyping the threat.
- Australia plans systems-based regulation for frontier AI, borrowing from workplace safety and banking supervision models.
- There are 109 machine identities for every human, and a Fortune 500 company found 10,000 hidden AI agents it did not know about.
- AWS reported AgentCore usage grew 15-fold in six months as AI agent adoption accelerates.
- A federal AI pilot program denies Medicare claims for seniors, with Virtix Health denying more claims than it approved for 17 procedures in early 2026.
- The AI cybersecurity market is expected to grow by nearly 25 percent annually until 2033.
- A 2026 national security exercise used Anthropic's Mythos AI model to identify government vulnerabilities in hours.
- Anthropic CEO Dario Amodei has shared views on how AI will reshape American workplaces.
- Law and Order's 26th season premiere explores risks of AI manipulating the justice system and highlights the need for human oversight.
- Researchers say reliable AI sea ice predictions depend on sustained, quality observational data across both poles.
AI executive Andy Barkett warns against overhyping extinction risk
Andy Barkett, a former Google and Nvidia engineering manager, says assigning a 10 percent chance to AI killing humanity is misleading. He believes AI could bring great prosperity or serious disruption, but we lack the history to predict outcomes. He worries more about AI making existing problems like misinformation and job loss worse. He also says government regulation should focus on accountability rather than slowing down progress.
Australia plans systems-based regulation for frontier AI
Australia is considering a systems-based approach to regulate frontier artificial intelligence. Assistant minister Andrew Charlton said companies would need to show they have rigorous processes for finding and managing AI risks. He argued voluntary codes fail and detailed rules become outdated quickly. This approach would borrow models from workplace safety and banking supervision.
SailPoint Navigate highlights identity security challenges for AI agents
At SailPoint's Navigate 2026 event, experts stressed that identity security for AI agents is now a top priority. There are 109 machine identities for every human, and agents can gain access to systems unintentionally. AWS reported AgentCore usage grew 15-fold in six months. A Fortune 500 company found 10,000 hidden agents it did not know about. The focus is on just-in-time access and human oversight.
AI sea ice forecasts need better long-term observations
Researchers say reliable AI-based sea ice predictions depend on sustained and quality observations across both poles. AI sea ice prediction systems have grown from almost none a decade ago to dozens today. They can now run in minutes on a laptop instead of days on a supercomputer. However, the data needed to test and improve these systems remains sparse. Both traditional models and AI methods rely on good observational data.
Data centres and AI become central to national defence and security
A report says AI and data centres are now on the frontline of national security. The market for AI in cybersecurity is expected to grow by nearly 25 percent each year until 2033. A 1997 US cyber exercise called Eligible Receiver led to the creation of US Cyber Command. A 2026 exercise reportedly used Anthropic's Mythos AI model to find government vulnerabilities in hours. Supply chains for semiconductors and critical materials have become key security concerns.
AI and Quantum Computing: Real Progress and Big Challenges
AI and quantum computing are often described as a perfect match, but the benefits are not as clear as some claim. Quantum computers may help train better AI models for chemistry and materials science. In a surprising twist, AI is already helping quantum computers work better by managing their hardware and fixing errors. The real connection between the two technologies is more practical than dramatic.
Federal AI Program Denies Care to Washington Seniors
A federal pilot program uses AI to review and deny Medicare claims for seniors in Washington and five other states. The program, run by companies like Virtix Health, is paid based on how many claims it denies. In early 2026, Virtix Health denied more claims than it approved for 17 procedures, including nerve stimulation and cervical fusion surgery. Doctors and lawmakers say the delays harm patients' health and quality of life.
AI Is Already Changing American Workplaces
AI is transforming how Americans do their jobs, with both positive and negative effects. Tech leaders have long predicted major workplace changes, but the exact form those changes will take is still uncertain. Anthropic CEO Dario Amodei has shared views on how AI will reshape work. The shift is already visible in many industries across the country.
Law and Order Explores Dangers of AI Power
The 26th season premiere of Law and Order shows a world where powerful AI systems can manipulate the justice system. The episode follows a young woman unfairly targeted by a compromised AI used by police. Writers use the story to warn about the risks of letting AI make decisions that affect people's lives. The episode also highlights the need for human oversight and the role of hackers trying to expose AI flaws.
Sources
- Will AI kill us all? Political Economy Alumnus and AI executive Andy
- Australia looks to systems-based regulation for frontier AI standards | MLex | Specialist news and analysis on legal risk and regulation
- Identity security for AI agents: 18 insights from Navigate 2026
- Reliable AI Sea Ice Forecasts Depend on Sustained Observations
- Defence goes digital – data centres and AI on the frontline of national security
- When AI Meets Quantum: The Next Great Leap in Computing
- The feds are using AI to deny care to WA seniors, Sen. Murray says
- Welcome to the AI Workplace
- Even ‘Law & Order’ Is Terrified of AI
Comments
Please log in to post a comment.