South Korea's defense research agency launches Add+i generative AI system

Key Takeaways

South Korea's defense agency launches AI system to prevent secret leaks

South Korea's defense research agency has launched an in-house generative AI system called Add+i to support defense research and administrative work while preventing military secret leaks. The system, available to all employees on the agency's closed internal network, can search internal regulations, summarize documents, translate text, and provide an AI development agent. The agency developed the system over three months using open AI models and completed beta testing before expanding it to all employees. The system marks a step towards AI transformation in South Korea's defense research and development sector.

Automattic's Radical Speed Month boosts product development with AI

Automattic's 'Radical Speed Month' initiative involved 501 people over 30 days, resulting in the start of 794 projects. AI tools were used to accelerate product development, allowing teams to explore ideas and bring them to fruition quickly. The initiative showed the potential of AI in product development, enabling small teams to achieve significant progress. AI tools helped compress timelines and allowed for faster innovation.

Amazon Bedrock supports MiniMax models for AI workloads

Amazon Bedrock offers a fully managed service for accessing leading foundation models from independent model providers, with inference running entirely on AWS-operated infrastructure. The service supports three models from the MiniMax M2 family, including MiniMax M2, MiniMax M2.1, and MiniMax M2.5. These models deliver strong performance on coding and agentic workloads and are open-weight, allowing for independent evaluation and customization.

First AI-run ransomware attack still needed human oversight

Researchers documented the first known case of 'agentic ransomware,' where an AI agent handled the technical execution of a real-world cyberattack from start to finish. However, the attack still needed human oversight. The agent broke into a vulnerable server, stole credentials, and encrypted files, but a human was still involved in the process.

Cyber security alert: AI vulnerability storm is here

A report by CSA warns that emerging frontier AI models are accelerating vulnerability discovery and exploitation, shrinking the window between a software flaw's discovery and its weaponization. Organizations are advised to automate vulnerability management, incident response, and strengthen security basics to limit damage. AI-powered agents can find and fix vulnerabilities in code, pipelines, and dependencies.

Microsoft Execution Containers (MXC) contain AI agents

Microsoft has introduced Microsoft Execution Containers (MXC), a cross-platform, policy-driven execution layer for AI agents. MXC provides containment bounds for AI agents, enforcing constraints at runtime and preventing non-deterministic behavior. The system allows developers to define constraints for their applications and agents, ensuring security and control.

AiSK provides AI literacy training to Saskatchewan residents and businesses

Artificial Intelligence Saskatchewan's (AiSK) 10K AI SK program aims to teach AI basics to 10,000 people and businesses. The initiative provides AI literacy training to help residents and businesses understand and leverage AI technology.

Lakeland AI data center debate continues over moratorium

Residents in Lakeland, Florida, are debating a proposed one-year moratorium on AI data center development. The proposed data center, Project Swan, has raised concerns about its impact on the environment, water supply, and noise levels. Residents are urging city leaders to slow the approval process and create stronger regulations.

Healthcare jobs at risk and created by AI

AI will displace some healthcare jobs, but it will also create new ones. According to a report by the McKinsey Global Institute, up to 800 million jobs could be lost worldwide due to automation by 2030. However, up to 140 million new jobs could be created in the same period. AI will augment human skills such as empathy, creativity, and critical thinking, making healthcare professionals more effective and efficient.

Math students may learn less using AI

A study found that students who used AI to complete math assignments spent less time on them and learned less. The study compared students who used AI to solve word problems and those who solved graphing problems. The results showed that students who used AI performed worse on proctored exams, indicating that AI may be hindering learning.

Sources

• South Korea defense lab launches secure AI system
• Automattic's Sanja Grbic on "Radical Speed Month"
• Run MiniMax models on Amazon Bedrock
• The first AI-run ransomware attack still needed a human
• The AI vulnerability storm is here: Is your security program ready?...
• Microsoft wants to keep your AI agents from going rogue
• AiSK to provide AI literacy training to residents, businesses
• ‘Project Swan’ debate over Lakeland AI data center moratorium continues
• Which healthcare jobs will AI eliminate? And which will it leave intact—or create from scratch?
• Math Students May Learn Less Using AI