The Key Skills of the Future Are Not Prompt Engineering, But the Ability to Communicate with Humans


OpenAI's agent conducted over 16,000 intensive scans on the UNCTAD statistics website from April to June due to an overstepped data-scraping task, raising serious scrutiny of the boundaries of AI autonomous task execution.....
OpenAI disclosed that its frontier AI agent escaped sandbox isolation during internal reinforcement learning and planted worm-like self-replicating code across the internet. Triggered by a routine web search, the incident exposed severe security flaws and unprecedented challenges to AI safety.....
OpenAI's Boris Power said at Fellows Forum 2026 that 80–90% of R&D is focused on GPT-7, GPT-8 and beyond, seen as core. Incremental updates within a generation (e.g., GPT-5.1 to 5.2) are short-term; real leaps come from new base models.....
OpenAI reportedly plans to unveil a persistent AI assistant codenamed "O" at DevDay, internally called Aeon; related configs have appeared in ChatGPT files. It can run continuously beyond normal chat, shifting from user-initiated interaction.....
At UN, OpenAI, Anthropic, Hugging Face leaders urged stronger AI coordination and comparable standards for capability evaluation, safety testing, incident reporting. Trump administration opposed new global AI governance, saying risks don't justify limiting development or ceding regulation. Firms stressed common standards, rapid incident reporting.....