Enterprise
OpenAI Raises $122 Billion in Record Round
OpenAI has secured $122 billion in funding to accelerate the next phase of AI development. The investment, announced on the company's official blog, represents the largest single fundraise in AI history and will fund infrastructure buildout and frontier model research. No specific valuation or lead investors were named in the initial announcement. Source →
Policy
Anthropic Signs AI Safety MOU With Australia
Anthropic and the Australian government have signed a Memorandum of Understanding focused on AI safety and research collaboration. The partnership aims to enhance safety practices and promote responsible AI development, marking one of the first formal bilateral agreements between a frontier AI lab and a national government. Details on specific research programs or funding commitments were not disclosed. Source →
Agents
Mistral Launches Spaces CLI for Agents
Mistral has released Spaces, a command-line interface designed for both human developers and AI agents. The company says the process of designing for agent workflows forced it to build better tooling for everyone, resulting in a CLI that streamlines user-agent interactions. Spaces represents Mistral's growing investment in agentic infrastructure beyond pure model releases. Source →
Policy
OpenAI Safety Bug Bounty Targets AI Misuse
OpenAI expanded its bug bounty program on March 26 to cover AI safety and misuse risks, moving beyond traditional software vulnerabilities. The new Safety Bug Bounty Program rewards researchers who identify ways models can be manipulated for harmful outputs. Alongside it, OpenAI published its Model Spec framework detailing principles for model behavior, instruction-following, and conflict resolution. Source →
Enterprise
Amazon and OpenAI Build Stateful AI Runtime
Amazon and OpenAI announced a partnership on March 26 to co-develop a Stateful Runtime Environment on Amazon Bedrock. The system emphasizes persistent memory and tool-use infrastructure, enabling AI agents to maintain context across sessions and interact with external services. Availability is planned for the coming months, positioning Bedrock as a key platform for agentic applications. Source →
Model Release
Google Ships Gemini 3 Deep Think Model
Google launched Gemini 3 Deep Think on March 26, targeting scientific and engineering tasks such as identifying logical flaws in math papers and optimizing fabrication methods. The model is live in the Gemini app for Ultra subscribers with early API access available. It launched alongside Lyria 3 and Lyria 3 Pro music generation models as part of Google's expanding specialized model family. Source →
Trend
Q1 2026 Release Pace Cools After Record Surge
After hitting 255 model releases in Q1 2026, the final days of March saw a notable slowdown with no major new model launches between March 27 and 31. The quiet stretch follows a concentrated burst of activity earlier in the month, including GPT-5.4 on March 5, Mistral Small 4 around March 17, and NVIDIA Nemotron 3 Super near March 11. The pattern suggests labs are shifting resources toward infrastructure and agentic tooling rather than raw model releases. Source →