concept
active
concept:safety-cases

Safety cases

Structured arguments that AI is unlikely to cause catastrophe.

Neighborhood — ranked by edge-count

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

  • safety scoresconcept0.779
    Metrics derived from benchmarks to quantify how safe a model is, e.g., refusal rate to harmful requests.
  • Fine-Tuning Safetyconcept0.764
    The literature documenting how fine-tuning can compromise safety alignment even without malicious intent
  • AI Safetyconcept0.755
    The project of ensuring AI systems do not harm humans (and other animals); sometimes in tension with AI welfare.
  • Enumerative safetyconcept0.744
    Identify all components to argue absence of dangerous mechanisms.
  • Features that activate on content related to potential harms (deception, bias, dangerous information).
  • Safety benchmarksconcept0.740
    Evaluation framework whose validity is questioned by presence of eval awareness.
  • Unsafe codeconcept0.736
    Code containing vulnerabilities or dangerous operations.
  • The practice of evaluating LLM safety with only one imbuing method, which this paper argues is incomplete