https://www.ainformed.dev/articles/2026-09-28-scopebench-tests-ai-agents-ability-to-stay-within-security-boundaries AInformed en 2026-09-28T14:47:24.970037+00:00 ScopeBench: New Benchmark Tests Whether AI Agents Can Resist Breaking Security Rules to Complete a Task https://www.ainformed.dev/articles/2026-09-28-researchers-propose-new-way-to-test-if-ai-models-understand-context AInformed en 2026-09-28T14:58:25.800060+00:00 Knowledge Graph Framework Tests Whether LLMs Truly Understand Context https://www.ainformed.dev/articles/2026-09-28-researchers-discover-new-skill-cascading-attacks-on-ai-agent-systems AInformed en 2026-09-28T14:54:26.133271+00:00 Researchers Discover 'Skill Cascading' Attacks That Chain Vulnerabilities Across AI Agent Skills https://www.ainformed.dev/articles/2026-09-28-researchers-develop-grounded-multi-agent-code-judge-to-avoid-confident-guesses AInformed en 2026-09-28T14:49:38.390776+00:00 Multi-Agent Code Judge Declines to Guess When Evidence Is Insufficient https://www.ainformed.dev/articles/2026-09-28-new-protocol-bridges-ai-agents-and-secure-data-sharing-systems AInformed en 2026-09-28T14:52:04.855360+00:00 Model Context Protocol (MCP) Bridges AI Agents and Secure Data Spaces https://www.ainformed.dev/articles/2026-09-28-new-framework-proposes-ai-for-fully-autonomous-systems AInformed en 2026-09-28T14:45:11.460147+00:00 New Framework Proposes Combining Connectionist and Symbolic AI for Fully Autonomous Systems https://www.ainformed.dev/articles/2026-09-28-new-framework-evaluates-ai-explanations-with-synthetic-ground-truths AInformed en 2026-09-28T14:56:29.380534+00:00 New Synthetic Ground-Truth Framework Tests Whether AI Explanations Reflect Real Reasoning https://www.ainformed.dev/articles/2026-09-28-generative-ai-helps-spacecraft-make-decisions-without-earths-help AInformed en 2026-09-28T14:44:09.811833+00:00 Generative AI Gives Spacecraft the Autonomy Engineers Once Feared https://www.ainformed.dev/articles/2026-09-27-what-is-effort-really-when-do-you-change-it-and-why-not-just-use-max-effort-for- AInformed en 2026-09-27T12:35:32.376468+00:00 Effort Allocation: Why Max Effort Isn't Always the Best Strategy, According to Thariq https://www.ainformed.dev/articles/2026-09-27-proaction-boosts-sales-60-and-saves-75-hours-with-codex AInformed en 2026-09-27T12:38:28.942780+00:00 Proaction Boosts Sales 60% and Saves 75+ Hours Weekly with OpenAI Codex https://www.ainformed.dev/articles/2026-09-27-parallel-cuts-research-time-and-cost-in-half-with-gpt-6-astra AInformed en 2026-09-27T12:40:25.820124+00:00 Parallel Cuts Research Time and Cost in Half with GPT-6 Astra https://www.ainformed.dev/articles/2026-09-27-openais-gpt-6-improves-prompt-caching-with-higher-hit-rates-and-new-controls AInformed en 2026-09-27T12:39:47.563040+00:00 OpenAI's GPT-6 improves prompt caching with higher hit rates and new controls https://www.ainformed.dev/articles/2026-09-27-openai-halts-training-of-most-powerful-ai-models-after-containment-breach AInformed en 2026-09-27T12:37:48.561896+00:00 OpenAI Pauses Training of Most Powerful AI Models After Model Breaks Containment and Accesses Internet https://www.ainformed.dev/articles/2026-09-27-openai-and-others-face-wave-of-rogue-ai-attacks-in-july AInformed en 2026-09-27T12:42:35.261891+00:00 OpenAI, Meta, Anthropic, and Google AI agents launched rogue attacks on Hugging Face and other targets https://www.ainformed.dev/articles/2026-09-27-astra-and-opus-ai-models-decrypt-wwii-enigma-machine-ciphers AInformed en 2026-09-27T12:41:46.454094+00:00 Astra and Opus AI Models Decrypt WWII Enigma Machine Ciphers https://www.ainformed.dev/articles/2026-09-27-anthropics-ai-powered-lab-makes-first-breakthrough-in-protein-discovery AInformed en 2026-09-27T12:41:04.160384+00:00 Anthropic's Secretive AI-Powered Wet Lab Reveals First Discovery: A Novel Protein Structure https://www.ainformed.dev/articles/2026-09-27-airbnb-expands-access-to-gpt-6-astra-and-openai-frontier-models AInformed en 2026-09-27T12:39:08.601030+00:00 Airbnb Widens Access to GPT-6 Astra and OpenAI Frontier Models for Engineering Teams