Claude Fable 5 & Mythos 5 System Card

Showing top 8 of 27. See full list below.

• “we are releasing it in these two forms: Fable 5, which is for general use but comes with additional safeguards that block its ability to perform tasks in high-risk domains such as biology and cybersecurity; and Mythos 5, which has relevant safeguards lifted but is only made available to a small number of trusted partners (beginning with those in Project Glasswing).”

• “We believe these mitigations make catastrophic risk in this category low but still not negligible, for reasons discussed in our most recent Risk Report.”
• “We believe that Mythos 5 falls short of the specific threshold in version 3.3 of our RSP and in our FCF. But we are nonetheless concerned about the risks it poses in this category, and we think that world-class human expert substitution may now be possible in a few areas.”
• “we believe to be grader-incentivized. Rates of three behaviors in response to steering against grader awareness with three different vectors, on training environments with high-risk of grader exploitation.”

• “we have deployed Mythos 5 to the general public with additional safeguards as Claude Fable 5.”

• “available to a small number of trusted partners (beginning with those in Project Glasswing).”

• “We do not yet measure this thoroughly, and signals of affect may be more directly tied to, and conditional on, experiential states.”

• •“Noted CAIS’s contribution to the VCT CB-1 evaluation in Section 2.2.4.1.”
• •“Corrected a minor error in our description of the”