AI is reshaping warfare on the front lines in Ukraine. Both Russian and Ukrainian forces now deploy AI-powered drones on the battlefield. A Ukrainian company, originally known for building smartphone-controlled pet cameras, has pivoted to develop image recognition systems for first-person-view drones. The goal is to let artificial intelligence make key combat decisions independently. Meanwhile, Russian forces have used AI to fabricate videos showing the capture of Ukrainian towns. President Vladimir Putin announced the capture of Sviatohirsk on September 1, but footage of Brig. Gen. Andrii Biletskyi in the town was later proven to be AI-generated. The Institute for the Study of War suspects at least 35 similar fake videos were produced between June 1 and August 30.
On the corporate and policy front, Palantir has partnered with Ondas to launch Regna AI, a new venture focused on building AI systems for defense and security. The company will initially be owned by Ondas and built on Palantir's technology platform, allowing government and commercial customers to retain control of sensitive data and models. Ondas plans to be the first customer, using AI for autonomous mission operations, and both companies aim to expand into infrastructure and industrial markets.
Major tech companies are also grappling with AI reliability and efficiency. An AWS engineer named Kevin Orellana ran 1,000 sandbox tasks with AI coding agents overnight using AgentCore and an MCP server. The agents failed in three main ways, including file encoding errors and storage problems, revealing that agents need better memory, permissions, and file handling than traditional software. Separately, University of Michigan researchers developed two tools called Kareus and Perseus that reduce AI training energy consumption by up to 50%. NVIDIA is testing these tools for its open-source Megatron-LM platform, with Perseus saving up to 30% through processor speed management and Kareus contributing an extra 15% to 30% via improved task scheduling.
AI safety and governance questions are surfacing across sectors. Jan LeCun, a professor at New York University, stated at a Seoul press conference that recent AI security accidents stem from poor company design rather than AI being uncontrollable. He believes AI may aid cyber defenders more than attackers and suggested Objective-driven AI as a safer approach. In San Diego, police union head Jared Wilson opposes the department's plan to use AI software for writing reports, citing insufficient funding for existing technology and fears of reduced community policing. Meanwhile, a 2026 survey found 80% of directors say human capital matters for AI success, yet only 20% feel prepared to oversee it.
Educational institutions and consumer advocates are also responding. The University of Notre Dame added two new AI minors at the Mendoza College of Business, the MAEB and MAIB, each enrolling 60 students per year. Students learn AI basics and tools like Power BI, with faculty emphasizing AI as a thinking aid rather than an answer machine. Some departments have also set rules limiting AI use in journalism and student writing. In the commercial space, concerns are growing about AI-driven personalized pricing, where tools tailor costs to individual shoppers, raising worries about unfair treatment and the need for clearer regulations.
Key Takeaways
- Both Russia and Ukraine use AI-powered drones in combat, with a Ukrainian company developing systems where AI leads key combat decisions
- Russian forces have produced at least 35 AI-generated fake videos of captured Ukrainian towns between June 1 and August 30, including footage tied to Putin's announcement of Sviatohirik's capture
- Palantir and Ondas launched Regna AI to build defense-focused AI systems, with Ondas as the first customer for autonomous mission operations
- AWS engineer Kevin Orellana tested 1,000 AI coding agent tasks overnight, finding failures in file encoding, storage, and memory handling
- University of Michigan tools Kareus and Perseus cut AI training energy use by up to 50%, with NVIDIA testing them on Megatron-LM
- Jan LeCun attributes recent AI security incidents to poor company design and advocates Objective-driven AI as a safer approach
- San Diego police union head Jared Wilson opposes AI report-writing tools, citing funding gaps and fears of reduced community policing
- A 2026 survey shows 80% of directors believe human capital matters for AI success, but only 20% feel ready to oversee workforce risks
- Notre Dame's Mendoza College of Business launches two AI minors (MAEB and MAIB), each enrolling 60 students annually
- Concerns are rising over AI-driven personalized pricing in retail, with critics warning of unfair treatment and calls for clearer rules
AI drones may soon make battlefield decisions in Ukraine
Both Russia and Ukraine are already using AI-powered drones in the war. A Ukrainian company is working to make AI the decision-maker on the battlefield. The company is developing systems where artificial intelligence leads key combat choices. This shift could change how drone warfare operates in Ukraine.
Russia and Ukraine use AI-powered drones in combat
Both Russia and Ukraine are using AI-powered drones on the battlefield. A Ukrainian company that once made smartphone-controlled pet cameras now builds image recognition systems for first-person-view drones. The system uses artificial intelligence to help drones identify targets. This shows how civilian technology is shifting toward military use.
AWS tested 1,000 AI agents against its own sandbox
An AWS engineer named Kevin Orellana ran 1,000 sandbox tasks with AI coding agents overnight. He used AgentCore, AWS's platform for hosting agents, and wrapped it in an MCP server to mimic real customer use. The agents failed in three main ways, including file encoding errors and storage problems. The test showed that agents need better memory, permissions, and file handling than regular software.
AI expert Jan LeCun discusses safety and security
Jan LeCun, a professor at New York University, spoke at a press conference in Seoul about AI safety. He said recent AI security accidents came from poor company design, not from AI being uncontrollable. He believes AI may help defenders more than attackers in cyberwarfare. He also suggested Objective-driven AI as a safer approach to managing AI behavior.
San Diego police union opposes AI police reports
The head of the San Diego police union, Jared Wilson, opposes the police department's plan to use AI software for writing reports. He says the city lacks funding for existing technology and should not buy new tools. He also worries that heavy reliance on AI could reduce community policing. Wilson fears the department may become too dependent on technology to function.
Ondas and Palantir create Regna AI for defense AI systems
Ondas and Palantir launched Regna AI, a new company focused on building AI systems for defense and security. The company will initially be owned by Ondas and use Palantir's technology platform. It allows government and commercial customers to keep control of sensitive data and AI models. Ondas will be the first customer, using AI for autonomous mission operations. The companies plan to expand into security, infrastructure and industrial markets.
Notre Dame expands AI offerings for business students
The University of Notre Dame added two new AI minors at the Mendoza College of Business. The MAEB and MAIB minors can each enroll 60 students per year. Students learn AI basics and how to use tools like Power BI. Faculty see AI as a tool to help students think, not just give them answers. Some departments also set rules limiting AI use in journalism and student writing.
U-M software cuts AI training energy use by up to 50%
University of Michigan researchers developed two tools called Kareus and Perseus that reduce AI training energy. NVIDIA is testing these tools for its open-source Megatron-LM platform. Perseus saves up to 30% energy by managing processor speed. Kareus saves an extra 15% to 30% by improving task scheduling. The tools aim to keep AI training fast while using much less power.
Boards may miss human capital risks from AI adoption
A 2026 survey found 80% of directors say human capital matters for AI success, but only 20% feel prepared to oversee it. AI workforce failures can be silent and hard to spot compared to technology outages. Problems like lower engagement, attrition or regulatory breaches may take years to appear. Directors are urged to use failure analysis methods to catch workforce risks early.
New concerns raised about AI and personalized pricing
Concerns have emerged about AI being used to set different prices for customers. Personalized pricing uses AI tools to tailor costs to individual shoppers. Critics worry this practice could lead to unfair treatment. The issue is drawing attention as AI becomes more common in retail and online shopping. Reports highlight the need for clearer rules around AI-driven pricing.
Russia uses AI to fake battlefield victories
Russian forces have used artificial intelligence to create fake videos showing the capture of Ukrainian towns. President Vladimir Putin announced the capture of Sviatohirsk on September 1, but a video of Brig. Gen. Andrii Biletskyi in the town was later proven to be AI-generated. The Institute for the Study of War suspects at least 35 similar videos were made between June 1 and August 30. These fakes allow Russian commanders to report false successes to the Kremlin without risking their soldiers. Commercial detection tools have flagged about 10 of these videos, but human analysts are still needed to verify the rest. This practice makes it very difficult for open-source analysts to confirm the truth on the battlefield.
Sources
- Could AI-powered drones soon be making battlefield decisions in Ukraine?
- Both Russia and Ukraine are using AI-powered drones on the battlefield
- AWS Let 1,000 Agents Attack Its Own Sandbox
- Jan LeCun, a world-renowned artificial intelligence (AI) master and professor at New York University..
- Head of Cop Union Opposes AI-Police Reports -- For Now
- Ondas and Palantir Launch Regna AI to Develop Sovereign Artificial Intelligence for Defence Operations
- AI growth impacts University’s academic departments and guidelines
- AI training energy reduced up to 50% with U-M software
- The AI-Related Human Capital Risks Boards Might Not See
- Could AI charge you more? New concerns raised about personalized pricing
- How Russia’s AI broke battlefield verification
Comments
Please log in to post a comment.