Wednesday, Sep 16, 2026
AI;DR
2026-09-16- OpenAI is backing a provision of the bipartisan FRONTIER Act in the House that would require leading AI companies to embed outside evaluators to verify their models’ safety. The legislative support follows OpenAI’s Sept. 12 commitment to give independent evaluators access similar to that of employees, matching a pledge by Anthropic.
- openaiOpenAI is expanding its advertising platform by allowing brands to sponsor custom AI agents within ChatGPT. This new feature enables companies to create specialized chatbots for specific tasks, such as product recommendations or customer service, which are then promoted to users interacting with the platform.
- llmA new visualization tool tracks the release dates and training cutoffs of 20 major AI models, revealing significant delays between when a model is trained and when it is made available to the public. This highlights the time lag in knowledge acquisition for AI systems.
- periodic-labsPeriodic Labs has developed a 1 trillion parameter open-source model, Neon, which outperforms general-purpose AI systems in quantum physics experiments. The model was trained on a massive dataset of superconducting qubit calibration data, demonstrating the potential for specialized AI in scientific research.
- googleGoogle has released two new conversational audio models, Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, which support real-time dialogue and background processing. The models are available for both consumers and developers, with the extended version offering higher precision for complex tasks.
- Elon Musk has proposed a new safety protocol where AI companies share their models with competitors for testing before public release. This
79 stories