OpenAI introduces a Jev-like API for fast AI decisions
OpenAI has introduced Decisions API, a limited-preview product that appears to offer functionality similar to Jev, a model released by TypeSafe AI for software automation. The service lets developers provide a fixed set of choices and receive probability-based outputs quickly and at low cost, according to the description of the product.
During OpenAI’s Dev Day, CEO Sam Altman said Decisions API could give the company’s Luna model predefined options, including image categories or possible agent behaviors. He said narrowing the model’s task could preserve image understanding, broad language support and safety protections while improving speed. TypeSafe CEO Diogo Almeida, a former OpenAI engineer, described the development on X as part of “the clone wars” and linked it to fast, intuitive “System One” processing.
The similarity between the products remains unclear because Decisions API has not been broadly tested by developers. One possible use is monitoring AI agents. Cybersecurity professional Shapor Naghibzadeh built a Jev-based demonstration that evaluates each agent action, blocking likely harmful steps and flagging uncertain ones. The article says this monitoring cost $2.94 with Jev, compared with $372 using a frontier LLM, in a test related to the Hugging Face incident.
Written by Sakhanda Wire AI from reporting by TechCrunch. Facts, quotes and full context are in the original article below.