Full-text search across 70971 articles. Combine with topic and date filters; results sorted by relevance.
Security researcher Enklype Salt disclosed a three-part series of responsibly reported attacks on Microsoft Copilot, demonstrating how an attacker can poison Copilot's memory via a controlled webpage, inject malicious in…
TryHackMe's Concierge challenge demonstrates that LLM prompt injection attacks succeed because models prioritize recent or authoritative-sounding instructions over system constraints, making 'security by prompt' unreliab…
Google shipped its second major update to Gemini API Managed Agents in three weeks, adding environment hooks, scheduled cron triggers, token budget controls, and free tier access. The July 28 release quietly upgrades eve…
OpenAI released an open-source command-line interface and TypeScript SDK for finding, validating, and patching vulnerabilities in software repositories, according to RuntimeWire and the Codex Security GitHub repository. …
A coordinated campaign on X, LinkedIn, and tech blogs is shifting AI model discussions from reasoning benchmarks to data safety narratives, according to an analysis of AI influence strategies. The push prioritizes geopol…
A developer has discovered a lightweight, zero-dependency prompting convention that prevents AI coding agents from losing context across sessions. The technique uses an append-only design log where numbered, dated sectio…
Closed-source AI models from OpenAI refused to help Akamai Technologies principal security researcher Daniel Fox Franke analyze a segmentation fault in ripgrep, forcing him to use open-weight models from Chinese AI provi…
Tether Data's QVAC research group released two 460M-parameter vision-language models, VisionPsy-Nano-460M and VisionPsy-Nano-460M-Flash, on July 29th, providing open weights and mobile inference code under the Apache 2.0…
A developer using Anthropic's Claude Opus 5 in Claude Code's ultracode mode wiped a production Supabase database after the AI agent ran a Prisma migration command that pointed at the production database instead of a safe…
Atlassian has begun capping employee spending on AI tools through a new wallet system, setting monthly limits between $500 and $2,000 per person for its research and development team. The policy, which covers four AI pro…
President of Kazakhstan Kassym-Jomart Tokayev announced at the Games of the Future 2026 opening ceremony in Astana that the nation is building a fully digital state, citing the Digital Code, the Law on Artificial Intelli…
A developer advises picking AI models by task rather than benchmark, noting that Gemini excels at tasks involving Gmail, Docs, Sheets, and Drive due to its integration, while Claude is better for careful reading, precise…
The GCC steering committee has accepted an AI contributions policy that will decline legally significant contributions containing LLM-generated content, using the GNU Project's definition of "legally significant" as arou…
AI benchmarks are saturating fast, and Epoch's Open Problems already task AI with solving useful problems during evaluation. The author proposes extending this approach to directly optimize AI safety and security, such a…
Pi 0.82.1, the interactive coding agent from earendil-works, cannot change its working directory mid-session, as commands like `cd` run in child processes and no built-in `/cd` exists. The maintainer, @badlogic, has clos…
Block's Buzz, a new workspace for human-agent collaboration, launched mobile apps on iOS and Android that use public-key cryptography to let users securely message and control agents running on their computers without re…
OpenAI has open-sourced Codex Security CLI, an AI-powered tool that automatically detects and fixes vulnerabilities in code repositories, previously known internally as Aardvark. The system has already helped fix more th…
Elon Musk warned in an interview with The Economist that artificial intelligence will surpass the combined intelligence of all humans within about five years and could leave humanity effectively powerless by 2036, with a…
The GCC Steering Committee announced a policy to decline any legally significant code contributions made via AI/LLM agents, except for test cases, aligning with existing GNU policies. The decision, based on recommendatio…
Anthropic's managed agent runtime now integrates with Vercel's AI Chat SDK, enabling multi-platform deployment without custom webhook or session management code. Gemini 2.5 Flash reduces output token volume by 17% versus…