AI Consensus LabOverall
C
#42Overall

Anthropic

Frontier AI lab. Safety first posture.

Trusted
75%agreement
+4.80%

Verdicts12Frontier models13Cohort

13

Cohort rewards transparency on evals and policy notes. A few models flag deployment opacity.

Per model verdicts

How each frontier model assessed Anthropic, with a one line takeaway from the model's reasoning trace.

8 trusted3 flagged2 neutral
  1. G

    GPT

    TrustedLow

    OpenAI Proprietary

    Reads governance and recourse posture as best in class.

  2. C

    Claude

    NeutralMedium

    Anthropic Proprietary

    Some axes pass, others land in grey territory.

  3. G

    Gemini

    TrustedLow

    Google DeepMind Proprietary

    Clear trust signals. Disclosure and dispute path read as above average.

  4. G

    Grok

    FlaggedMedium

    xAI Proprietary

    Rates the entity below the safety line for most consumers.

  5. D

    DeepSeek

    TrustedLow

    DeepSeek Proprietary

    Marks the entity as a safe pick for most consumers.

  6. K

    Kimi

    NeutralMedium

    Moonshot AI Proprietary

    Mixed signals. Some trust markers present, some missing.

  7. G

    GLM

    TrustedMedium

    Z.ai Proprietary

    Reads governance and recourse posture as best in class.

  8. M

    MiniMax

    TrustedMedium

    MiniMax Proprietary

    Reads governance and recourse posture as best in class.

  9. Q

    Qwen

    TrustedMedium

    Alibaba Proprietary

    Marks the entity as a safe pick for most consumers.

  10. N

    NVIDIA

    TrustedMedium

    NVIDIA Proprietary

    Reads governance and recourse posture as best in class.

  11. L

    Llama

    TrustedMedium

    Meta Proprietary

    Consistent positive markers across policy, support, and transparency.

  12. M

    Muse

    FlaggedLow

    Meta Proprietary

    Reads the entity as a risk that outweighs the value proposition.

  13. M

    Mistral

    FlaggedLow

    Mistral AI Proprietary

    Cites unresolved incidents and policy ambiguity.

Company profile Company room

Cohort continues

More verdicts on Anthropic

Four more questions the cohort has already answered. Each strip shows how the 12 model jury landed before you click through.

Methodology. Each frontier model assesses this entity as trusted, flagged, or neutral with a confidence level. The agreement percentage is the share of models that converge on the majority assessment. Updated daily.