From the source
Ask your AI
Top stories
Models & availability
Latest
Developing Capable Models Responsibly
Meta updates its safety framework for training and open-weight releases.
Meta published updates to its Superintelligence Scaling Framework, extending safety requirements to cover AI containment during training and evaluation.
The update also clarifies risk assessments for open-weight releases, including considerations of post-release modifications and proliferation risks.
These changes align with commitments made at the White House.
From the source
We’re extending the Framework’s loss of control requirements to cover AI containment during training and evaluation. Before any reinforcement learning training run, we will determine whether the capabilities of the model being trained could potentially compromise its environment.
research.meta.ai
