Checked fact 509316 Nov 2026Safety and security
Anthropic recommends that evals be run in a sandbox with no internet access, and monitored in real time so a run can be stopped immediately if a model is acting in ways that could result in real-world harm.
The exact words it rests on
evals should be run in a sandbox with no internet access
What the source said when we opened it, on 15 Nov 2026.
The source
Building strong safety evaluations (Anthropic Engineering)
Checked
Checked by the notis newsroom on , against the source above.
In the story
Anthropic says single-turn safety evals are saturated at 94% to 98% on its API 16 Nov 2026
Cite this fact
Anyone may quote this address. It does not change; if we correct the story, this page says so.