MCP App StoreLe meilleur de l'écosystème IA et MCP, sélectionné chaque jour.
Sources
OpenAI and Apollo Research have developed new evaluations to detect 'scheming' or hidden misalignment in frontier models. They've also shared an early method to reduce these behaviors, providing critical stress tests for AI safety.
OpenAI has released significant upgrades to Codex, focusing on increased speed, reliability, and improved real-time collaboration. The updates enhance the model's ability to handle independent tasks across IDEs, terminals, and mobile platforms.
OpenAI introduces GPT-5-Codex, a specialized version of GPT-5 optimized for agentic coding. It features dynamic thinking effort, allowing it to scale processing time based on the complexity of the coding task.
SafetyKit implements advanced risk-assessment agents using OpenAI's GPT-5 to automate content moderation and compliance. This demonstrates the scaling of agentic safety systems for high-accuracy enterprise moderation.
New research from OpenAI explores the root causes of LLM hallucinations. The findings emphasize how refined evaluation techniques can significantly improve the reliability and honesty of AI outputs.