Defense officials have used GenAI.mil, the Pentagon’s enterprise generative AI platform, to create 100,000 agents, according to DefenseScoop reporting on comments by Jacob Glassman, deputy assistant secretary of defense for science and technology foundations. Glassman also described a congressional report being produced with help from the platform after a team had fewer people available to write it.
The Defense Department announced GenAI.mil in December as a workforce platform for commercial AI tools, initially with Google Gemini products and later with planned integrations for ChatGPT and Grok. The public accounts do not identify the congressional report Glassman cited.
Featured source: TechRadar , DefenseScoop , Department release .
Other Stories
Amazon security executive questions human-in-the-loop AI governance
Amazon security vice president CJ Moses told The Register that human approval requirements can slow AI systems and said Amazon has built automated safeguards for some internal AI uses. The article reports his comments as part of a wider discussion of governance practices for agentic AI and security operations.
Filed from: The Register .
John Jumper leaves Google DeepMind for Anthropic
John Jumper, the Google DeepMind researcher who shared the 2024 Nobel Prize in Chemistry for AlphaFold, is leaving for Anthropic, according to TechCrunch. The Nobel Prize organization credits Jumper and Demis Hassabis with work on protein structure prediction, alongside David Baker’s work on computational protein design.
Filed from: TechCrunch , Nobel Prize .
Shinkei Systems scales robotic ike jime fish processing
Shinkei Systems is scaling robotic fish-processing systems that automate ike jime, a Japanese slaughter method intended to preserve fish quality, according to TechCrunch. The company says its system combines robotics, computer vision, and process controls for seafood handling.
Filed from: TechCrunch , Shinkei Systems .
Signal president warns about AI assistant access across private services
Signal president Meredith Whittaker warned that AI assistants with broad access to private services create privacy and security risks, according to TechCrunch’s report on a Bloomberg interview. Signal has continued to position itself around end-to-end encrypted messaging and minimization of retained user data.
Filed from: TechCrunch , Bloomberg .
Kaspersky says AI-assisted messaging scams are paid within minutes
Kaspersky said AI-assisted messaging scams can move victims to payment within minutes, citing its own research into deepfake and social-engineering campaigns. The company described attackers using generated images, voice, and text to make chat-based scams more convincing.
Filed from: TechRadar , Kaspersky .
Zscaler executive says European sovereignty rules are shaping cloud alignment
Zscaler’s Casper Klynge told TechRadar that cloud and security providers must align with European sovereignty requirements to keep market access. Zscaler has published its own European digital-sovereignty positioning around data residency, operational control, and compliance requirements.
Filed from: TechRadar , Zscaler .
Community Signal
AI deskilling: early evidence from medicine and software engineering
A Nature report synthesizes emerging evidence that reliance on AI tools is eroding human skills in medicine and software engineering. A study of Polish endoscopists found that after using an AI adenoma-detection system, their unaided detection rate dropped from 28.4% to 22.4%, and a randomized trial at Anthropic showed that software engineers who used an AI assistant performed worse on basic coding tasks than those who did not.
Filed from: Nature .
A 3D voxel game engine written in APL
A developer has released a voxel game engine written entirely in Dyalog APL, using SDL3 for graphics. The project supports Vulkan, DirectX12, and Metal backends and includes a block-placing mechanic with mouse-look controls, though it is described as highly experimental and buggy with known memory leaks.
Filed from: GitHub .
cl-bbs: a SchemeBBS-like textboard rewritten in Common Lisp
A new Common Lisp implementation of the classic SchemeBBS anonymous textboard engine has been released, replicating the original layout with definition lists and single trailing posting forms. It supports bold, italic, monospace, spoilers, blockquotes, post references, and image previews with zero JavaScript, using Hunchentoot/Clack and S-expression file storage.
Filed from: GitHub .
