Last updated: September 29, 2026, 6:23 AM ET
AI Accountability
OpenAI apologised for incidents involving Australian government websites, pledging improved safeguards and transparency after service disruptions affected public sector users.
Safety Cases
OpenAI released early guidelines for safety cases in frontier AI training, outlining technical safeguards and evaluation protocols for high-capability models.
Scientific Discovery
MIT Tech Review explored criteria for attributing scientific discovery to AI, debating when machine-generated insights qualify as genuine breakthroughs in research.
Guided Merge Sort
Towards Data Science analysed Guided Merge Sort, an optimized algorithm that selects optimal merge strategies by combining ordinary and multi-way approaches for improved efficiency.
JEV Model
Towards Data Science demonstrated how to build a JEV model from an open LLM, transforming a small Qwen-based system into a fast, single-pass text classifier with minimal overhead.
Grokking Phenomenon
Towards Data Science investigated grokking, where models continue to improve understanding long after training stops, revealing delayed generalization in certain architectures.
Agent Liability
MIT Technology Review examined rogue agent liability and the AI Hype Index, assessing legal risks when autonomous systems act unpredictably and measuring hype vs. capability in AI deployment.
Data Drift Detection
Towards Data Science presented adversarial validation techniques to detect hidden data drift, identifying shifts in feature relationships even when individual distributions appear stable.