Written Testimony of Marius Hobbhahn

The hearing is listed under the committee's subcommittee pages. The excerpt available from that page consists of a statement about the timing of independent safety evaluations of frontier AI models.

According to the excerpt, under the current testing regime, independent safety evaluations of frontier AI models typically happen in the weeks leading up to a model's public release. The excerpt does not identify which models, which testing regime, or which organisations conduct the evaluations. It does not state a specific number of weeks, name a regulator or evaluation body, or describe a proposal to change the current arrangement. The statement is limited to characterising when such evaluations typically occur relative to a model's public release.

It also does not indicate whether the testimony addresses other witnesses' evidence or the committee's questions. No further detail about the testimony's argument appears in the excerpt provided. On the excerpt provided, the testimony's reported content is the observation that independent safety evaluations take place in the period shortly before a frontier model is released publicly.