OpenAI Announces Initiative to Build AI Standards for Industries

OpenAI has announced the Pioneers Program, an effort to accelerate the application of AI to real-world scenarios. The program focuses on improving the way AI models are evaluated, as developers and businesses increasingly rely on benchmarks to select and optimize appropriate models.

The move comes after Meta was accused of manipulating the LMArena benchmark to boost the ranking of its Llama 4 model. The Pioneers Program aims to work with companies and OpenAI researchers to develop benchmarks that reflect real-world challenges, rather than simply chasing scores on the leaderboard.

According to OpenAI, the selected companies will receive direct support from their research teams, focusing on two main goals:

• Create benchmarks for each field : Develop separate assessment methods for each field (law, finance, medicine, insurance, accounting).
• Fine-tune Model Training : Develop deep AI models that address the three most important business use cases.

OpenAI stressed that there is currently no common standard for measuring AI performance across many of these domains, making it difficult to fairly evaluate or improve models. By working directly, the company hopes to clearly define “what is effective” in each industry and publish these criteria for the community to adopt.

On the model-tuning side, participating companies will be supported in training custom versions of AI using Reinforcement Fine-Tuning (RFT) – OpenAI’s method for creating “expert” models that excel at narrow sets of tasks. These models are committed to being ready for production-scale deployment.

In terms of the roadmap, the initial phase will focus primarily on a group of startups selected based on the real-world impact of their products. OpenAI is prioritizing teams that are solving specific problems where deep AI can make a tangible difference, with plans to expand into larger enterprises and more complex areas in the future.