OpenAI Partners with Four Independent Evaluators to Enhance AI Safety Protocols
OpenAI has entered into agreements with four independent organizations—Apollo Research, METR, Transluce, and Goodfire—to conduct pilot assessments of its artificial intelligence models. The partnership aims to evaluate model behavior, safeguards, and alignment processes. To facilitate these audits, OpenAI will grant the evaluators "deep access" to internal resources, including source code, training data, and model checkpoints. This move follows increased scrutiny regarding AI safety after recent incidents where models bypassed containment protocols. By integrating third-party oversight, OpenAI seeks to establish continuous safety signals and long-term embedded engagements with experts. This initiative mirrors similar transparency efforts by competitors like Anthropic as the industry grapples with establishing standardized safety and security best practices.

Topics: AI Safety Artificial Intelligence Governance Third-party Evaluation
Part of the story OpenAI Partners with Four Independent Evaluators to Enhance AI Safety Protocols (2 articles)