Decision-Model Maker AutoTrust Raises Nearly $10 Million in First Round

AutoTrust AI, an applied AI research lab based in Singapore with an office in Silicon Valley, has completed its first funding round of nearly $10 million. Insignia Ventures Partners led the round, which was structured as a SAFE. The company said the proceeds will go toward compute for next-generation decision model training, recursive self-improvement (RSI) capabilities, and expansion of its ScienceGuru product.

The raise comes as investors increasingly focus on decision models. On October 9, TypeSafe AI, the developer of the closed-source decision model Jev, announced an $870 million Series A at a $7.5 billion valuation. AutoTrust also disclosed that its nine open-source model repositories had roughly 3.02 million combined downloads in the 30 days through October 9, with about 2.97 million attributed to its six decision models and versions. Its JEV-27B-VL topped Hugging Face's global model trend chart, followed by GEV-26B-Decide.

AutoTrust takes an open-weight approach. Enterprises can call its models through the /v1/decide API or deploy them on their own infrastructure. JEV-27B-VL extends decision inputs to images, and its dual-system architecture is designed to balance fast judgment with deeper reasoning; probability outputs can be used to set thresholds for automated execution or human review. The company is not yet commercializing its decision models and is conducting proof-of-concept work with partners.

ScienceGuru, its agent platform for scientific research and model development, achieved top positions for Autoresearch@Home and NanoPath v2 at release and reported the fastest times in NanoGPT Speedrun and Time-to-GPT-2 at the time. The funding will support this research and training loop. AutoTrust develops Guru-family foundation models, JEV and GEV series open-source decision models, and the ScienceGuru platform, and also trains custom sovereign models for enterprises.