AI News — Wednesday, March 18, 2026
OpenAI announced the release of GPT-5.4 mini and nano, new smaller, more efficient models designed for intelligence at scale.
Nvidia's CEO Jensen Huang projected massive sales for its new Blackwell and Vera Rubin AI chips, indicating a trillion-dollar market opportunity.
Anthropic introduced Cowork, a new Claude Desktop agent designed to interact with user files and assist with tasks without requiring coding.
Mistral AI is focusing on providing tools for enterprises to build custom AI solutions, positioning itself as a key competitor to OpenAI and Anthropic.
Research demonstrates that AI models can be trained to recognize and apply 'scientific taste,' potentially accelerating scientific discovery.
Sourcetable launched an AI-powered spreadsheet and data platform designed to simplify data analysis and management through intelligent automation.
Innate introduced home robots that can be programmed as easily as AI agents, aiming to make robotics more accessible for everyday use.
OpenSeeker aims to democratize advanced search agents by open-sourcing their training data, fostering transparency and collaborative development.
Google announced Gemini 3.1 Flash-Lite, a new model optimized for efficient, scalable intelligence across various applications.
OpenAI detailed new methods and research for designing AI agents that are more resilient to prompt injection attacks, enhancing security.
OpenAI released 'symphony,' an open-source orchestrator designed to manage and coordinate autonomous development agents.
Railway raised $100 million to develop AI-native cloud infrastructure, aiming to compete with major cloud providers like AWS.
Nous Research released NousCoder-14B, an open-source coding model that offers a free alternative to commercial coding assistants like Claude Code.
Researchers introduced ILION, a framework for deterministic pre-execution safety gates to enhance the reliability and safety of agentic AI systems.
A study revealed widespread gender and pronoun biases in the moral judgments made by various large language models, highlighting ethical concerns.