Mathematician Levent Alpöge says he used Anthropic’s Claude to help prove the 167-year-old generalized Riemann hypothesis. The argument is written in Lean, an open-source language that checks mathematical proofs, and experts expect the formal result to hold—but they still need time to examine it and explain it in accessible terms. A proof would not give hackers a direct shortcut to breaking RSA encryption: testing whether a number is prime is different from factoring large numbers. Its broader significance may be evidence that AI can help produce new mathematical knowledge, though the implications for other fields and whether people can understand such results remain open questions. Philosopher Thomas Nagel, meanwhile, argues that today’s AI—including Claude—does not have consciousness, beliefs, or a mental life, even when it convincingly imitates them.
At a NATO conference in Izmir, Turkey, military leaders from the alliance’s 32 members and Ukraine discussed AI-driven warfare. NATO Adm. Pierre Vandier said forces must connect information and make decisions faster; on the front line, teams may have only a minute between a sensor alert and an enemy strike. Ukraine’s Oksana Ferchuk said solutions must reach troops within weeks and improve through testing and feedback. Speakers also stressed adapting drones to electronic warfare, connecting technology companies more closely to battlefield needs, and making older systems work with AI.
Businesses are also considering how AI agents could change shopping and selling. At VSMCamp & CSMOSummit 2026, experts said that as consumers delegate product discovery and comparison to agents, companies may need to rethink brands, data, and marketing and sales operations. They called for ways to verify brand information and clear rules, controls, and human responsibility when agents make purchasing decisions. Separately, MIT researcher Jim Glass discussed multimodal models that connect text, images, video, and audio, with potential uses in assistants and in healthcare, finance, and transportation. His research aims to help models learn from experience and adapt, while recognizing the challenges of building more human-like AI.
A warning about AI-enabled attacks says banks may be underprotecting databases even as they focus on models and applications. Mythos-class systems could scan networks, chain vulnerabilities into attack paths, and act at machine speed; malicious database changes may appear legitimate while putting financial records and compliance at risk. The article presents Liquibase Secure as a way to govern and record database changes across systems including Oracle, SQL Server, PostgreSQL, Snowflake, and Databricks. Separately, letters about North Carolina data-center oversight describe Pitt County commissioners voting on Nov. 16, 2026, to pause construction of an AI data center for two years while rules are considered. Residents raised concerns about electricity and water use, utility rates, noise, environmental effects, and regulation; the pause could end earlier if new rules are adopted.
In healthcare, JPMorgan’s Chris Schott said cancer treatments such as antibody-drug conjugates, bispecifics, and personalized vaccines are improving care, and newer therapies could displace Keytruda in five to seven years. AI is already helping speed clinical-data analysis and drug filings, but its use in early target discovery remains unproven and may take three to five years to evaluate. Schott put U.S. GLP-1 use in the high single digits and international use at about 2%; prices fell roughly 20% in 2026, and seniors can get supplies for $50 a month. Medical-device companies face lower estimates, co-pay pressure, and reduced procedure volumes, with a possible recovery in 2027.
Early investment in AI and legal technology is helping some law firms gain ground, according to Chambers’ Lisa Hart Shepherd. Latham & Watkins, Kirkland & Ellis, Goodwin, and White & Case rank in the top tier in both the UK and U.S. Chambers Advantage Index. U.S. firms have expanded in London through lateral hiring, while mergers and targeted investment have also supported growth. U.S. accelerators grew premium practices 78% faster and UK accelerators 54% faster. In the UK, legal technology and service hubs have helped firms build data capabilities and improve efficiency; Davis Polk and Macfarlanes have advanced through focused practice strengths.
Key Takeaways
- Levent Alpöge says Anthropic’s Claude helped him produce a Lean-checked proof of the generalized Riemann hypothesis; mathematicians still need to study and explain it.
- The claimed Riemann result would not directly enable attackers to break RSA by providing a shortcut to factoring large numbers.
- NATO and Ukrainian officials say AI-enabled warfare demands faster decisions, rapid delivery of battlefield solutions, adaptable drones, and integration with older systems.
- AI agents could reshape how customers find and buy products, prompting calls for verified brand data, clear controls, and human accountability.
- MIT researcher Jim Glass studies models that connect text, images, video, and audio, with potential applications in healthcare, finance, and transportation.
- A database-security warning highlights risks from AI-enabled attacks; Liquibase Secure is presented as a change-control and audit tool for platforms including Snowflake and Databricks.
- North Carolina letters report Pitt County’s Nov. 16, 2026, two-year pause on an AI data-center project while officials consider rules for large facilities.
- JPMorgan’s Chris Schott says AI already speeds clinical-data analysis and drug filings, while early AI-based drug-target discovery remains unproven; he cites GLP-1 price declines of about 20% in 2026.
- Medical-device companies face pressure on estimates, co-pays, and procedure volumes, with a possible recovery in 2027; newer cancer treatments could challenge Keytruda in five to seven years.
- Chambers says early AI and legal-technology investment is helping leading law firms; U.S. and UK accelerators grew premium practices 78% and 54% faster, respectively.
A claimed Riemann proof could signal a new role for AI in science
Mathematician Levent Alpöge says he and Anthropic’s Claude helped prove the generalized Riemann hypothesis, a claim that still needs verification. A proof would not directly give hackers a shortcut to break RSA encryption, because testing whether a number is prime is different from finding the factors of large numbers. The claim’s biggest impact could be evidence that AI can create and prove new knowledge, not just assist people with existing mathematics. That ability could eventually help researchers in physics, engineering and computer science, though it remains unclear whether people could understand the results.
Mathematician and Claude claim a solution to the Riemann hypothesis
Levent Alpöge says he worked with Anthropic’s Claude on a proof of the 167 year old Riemann hypothesis, one of mathematics’ most important unsolved problems. The result is written in Lean, an open source language that helps check mathematical proofs, and many experts expect the formal result to hold. However, mathematicians still need time to understand the proof and explain it in a way people can follow. Alpöge said he plans to help make the argument understandable after releasing it.
NATO races to keep pace with AI driven warfare
Military leaders and defense experts from NATO’s 32 member states and Ukraine met in Izmir, Turkey, for a three day conference on critical technology and digital warfare. NATO’s Adm. Pierre Vandier said success depends on connecting information and making decisions faster, as AI speeds up each step. On the front line, teams may have only one minute between a sensor alert and an enemy strike, while drones can become outdated if they do not adapt to electronic warfare. Ukraine’s digital transformation minister Oksana Ferchuk said solutions must reach troops within weeks and keep improving through testing and feedback. Defense leaders also said companies need closer ties to the battlefield and older systems must work with AI.
Drug advances offer hope as medical device companies face pressure
Chris Schott of JPMorgan said new cancer treatments, including antibody drug conjugates, bispecifics and personalized vaccines, are improving care, while better therapies could replace Keytruda in five to seven years. Robbie Marcus said medical device companies face falling estimates, pressure on co-pays and lower procedure volumes, with a recovery possibly arriving in 2027. Schott said AI already adds value by speeding up clinical data analysis and drug filings, but AI based early target discovery remains unproven and may take three to five years to assess. He also said US GLP 1 use remains in the high single digits and international use is about 2%, while prices fell about 20% in 2026 and seniors can get supplies for $50 a month.
Philosopher Thomas Nagel says today’s AI is not conscious
Philosopher Thomas Nagel told Wired that today’s AI systems are not conscious, despite how lifelike they seem. He argues that consciousness likely depends on physical processes linked to an organism and a nervous system, not just on how a system functions. Nagel said chatbots such as Anthropic’s Claude can imitate beliefs and morality but do not have real beliefs or mental lives. He also said people treating chatbots as conscious does not prove that they are.
AI agents push businesses to rethink marketing and sales
At VSMCamp & CSMOSummit 2026, experts discussed how AI agents may change the way people find, compare, and choose products. As consumers hand some of these tasks to AI, businesses may need to rethink their brands, data, and marketing and sales operations. The discussions also covered how to verify brand information and set limits, controls, and clear responsibility when AI takes part in purchasing decisions. Experts said businesses need connected systems and clear rules for human oversight as AI agents gain more abilities.
Jim Glass explains how AI models can connect text, images, video, and audio
Jim Glass, a researcher with the MIT-IBM Computing Research Lab and MIT’s Computer Science and Artificial Intelligence Laboratory, discussed multimodal foundation AI models in a Q&A. His work uses machine learning, AI, and algorithms to help models connect text, images, video, and audio. The research aims to help AI learn from experience, adjust to new situations, and make decisions using complex information. Possible uses include virtual assistants and chatbots, as well as AI systems in healthcare, finance, and transportation. Glass also discussed the challenges and limits of building more human-like AI.
Banks face growing risks from AI attacks on databases
The article warns that Mythos-class AI systems can scan business networks, find weaknesses, link them into attack paths, and act at machine speed. It argues that banks focus too much on AI models and applications while leaving databases, where decisions and records are stored, less protected. Attackers could alter database structures, transactions, or records in ways that look legitimate, threatening financial information and compliance with SOX, PCI DSS, SOC 2, and DORA. The article says manual tickets and scripts may not prove who made changes or under what controls. It presents Liquibase Secure as a database control system that tracks changes, applies policies, and creates tamper-evident audit evidence across systems including Oracle, SQL Server, PostgreSQL, Snowflake, DynamoDB, and Databricks.
Early AI investments help leading UK law firms pull ahead
Chambers’ Lisa Hart Shepherd says law firms that made major investments and adopted AI early are gaining ground in the UK and US. Latham & Watkins, Kirkland & Ellis, Goodwin, and White & Case reached the top tier in both markets in the Chambers Advantage Index. US firms have grown in London through lateral hiring, while mergers and focused investment have also helped firms expand. US accelerators grew premium practices 78% faster, and UK accelerators grew them 54% faster. In the UK, early investment in legal technology and service hubs helped some firms build data and improve efficiency, while firms such as Davis Polk and Macfarlanes have advanced through focused practice strengths.
Letters debate AI data centers and local oversight in North Carolina
The article presents letters from Florian Deltgen about AI and Joe Angelucci about data center rules in Chowan County. Deltgen says the Pitt County Commission voted on November 16, 2026, to pause construction of an AI data center for two years, while new rules are considered. He argues that China supplies many materials and components used in solar, wind, energy storage, and data center infrastructure, and says large AI data centers need major amounts of electricity. Pitt County staff identified gaps in rules for large facilities, including standards for power, water use, noise, and generator emissions. The pause followed residents’ concerns about utility use, possible rate increases, noise, environmental effects, and the county’s ability to regulate such projects; it could end sooner if new rules are adopted.
Sources
- What the Riemann hypothesis proof really means about AI, and the future
- AI claimed a solution to the 167 year-old Riemann Hypothesis — and a human did, too
- Fast and furious: NATO wrestles with speed of AI-driven warfare
- Pharma's patent pain is giving way to innovation
- Does Consciousness Philosopher Thomas Nagel Think AI Meets the Threshold?
- AI Agent places new demands on marketing and sales strategies.
- Q&A: Enabling more human-like interactions and understanding in multimodal foundation models
- AI-Driven Attacks on Banking Databases: Governance at Scale
- Big Bets, Early AI Investments Fuel the Fastest-Growing UK Firms
- Letters: Deltgen on Artificial Intelligence; Angelucci Says Chowan County Must Close Data Center Loopholes
Comments
Please log in to post a comment.