Reflection AI on AI labs should conduct risk assessments…

Reflection AI

AI research company developing open-weight frontier models and publishing an Open Safety Framework.

Safety is a lifecycle commitment, not a single checkpoint. Reflection assesses systemic risk before a model is released and stays engaged with it across the model’s life. [...] Internal use: The same systemic-risk assessment applies to internal use of frontier models, including research, operational systems, and internal products. Internal deployment is assessed against the same capability thresholds and safeguards as pre-release deployment.

Unverified (Oct 5, 2026)

Source:

Reflection AI, “Open Safety Framework,” published with its Beam model announcement on October 5, 2026. The company’s official framework says it assesses systemic risk before model release and applies the same risk assessment, capability thresholds, and safeguards to internal use and internal deployment of frontier models.