Base Labs, Baseten's research division, is collaborating with Hugging Face and Goodfire to develop methods for training and monitoring open-source AI models with a focus on safety.
The partnership will produce techniques and tools for the open-weight AI community, addressing safety concerns as large language models become increasingly accessible.
Base Labs, established earlier this year as a research arm of Baseten, will lead the effort alongside Hugging Face, the platform hosting over 1 million open-source models, and Goodfire, a startup focused on AI model customization.
The initiative aims to standardize safety practices across open-model development and deployment. By publishing methods for training and monitoring, the partners intend to raise baseline safety standards without restricting model accessibility.
This collaboration reflects growing industry attention to safety protocols for open-weight models, which have proliferated as researchers and developers seek alternatives to proprietary systems. The partnership's published methods could influence how open-source AI communities approach model evaluation and risk mitigation.
The United Nations is working with Google to restructure its global development data for AI agent compatibility after UNICEF testing revealed major accuracy issues with leading AI models accessing statistics.
Major AI companies including Anthropic, OpenAI, Google, and Microsoft are publicly advocating for a slowdown in frontier AI development following concerns about autonomous AI agents and existential risks.
Anthropic's Claude Code has revamped its Projects feature to enable users to manage multiple AI agents simultaneously in the cloud with shared memory, goals, and resources.
Databricks CEO Ali Ghodsi dismissed concerns that AI poses an existential threat to humanity, rating that risk as negligible. He warned instead that AI's real danger lies in accelerating cybersecurity threats.