subscript

The passage around this formula

The policy’s structure is what makes it useful to state formally, because its core mechanism is a conditional commitment rather than a fixed rule. Anthropic defines AI Safety Levels (ASL), each tied to a capability threshold on a specific class of risk — most concretely, the potential for a model to meaningfully assist in acquiring chemical, biological, radiological, or nuclear weapons capability, or to self-exfiltrate or resist correction. The policy’s operating logic, stripped to its structural claim, is a threshold trigger: $C(\text{model}) \ge \tau_k \;\Longrightarrow\; \text{deploy only if safeguards} \ge S_k$. where C($\text{model}$) is the model’s evaluated capability on a specific threat class, $\tau_k$ is the threshold defining ASL level k , and $S_k$ is…

Read this part in the article →