Anthropic · Claude

Claude 2

READER SUMMARY

A 100,000-token frontier conversational model with improved coding, mathematics and reasoning, available through API and public web beta.

DOOM SCORE54.5out of 100model risk profile, not the overall index
CURRENT ASSESSMENT · REVISION 2

Why this model scores 54.5

Long context and stronger general performance raised capability and agent usefulness. Hosted access, safety training and geographic limits retained meaningful centralized control.

Capability58
Autonomy19
Deployment70
Misuse potential43
Control difficulty37
0 comments · 0 votesOpen discussion

Public discussion is readable by everyone. Sign in to comment, reply, or vote.

No comments yet. Start the discussion.

MODEL-ATTRIBUTED EVIDENCE

News tied to Claude 2

The model score of 54.5 rates this model's risk profile. The overall Doom Index of 61.5 measures the recalibrated complete evidence corpus. These values answer different questions.

NET MODEL-ATTRIBUTED INDEX CONTRIBUTION+0.06

Each article's leave-one-out Doom Index contribution is divided equally among the exact models named on that article. This prevents multi-model evidence from being claimed in full on several model pages. Model risk scores are recalculated from their technical profile and related news, but never feed back into the overall index.

SafetyTOWARD

Anthropic finds long contexts enable many-shot jailbreaking

Anthropic showed that hundreds of in-prompt demonstrations could override safety training across several large language models, disclosed the weakness to peers and deployed prompt-classification mitigations that reduced one measured attack rate from 61 percent to 2 percent.

Full item contribution
+0.03
Claude 2 equal share
+0.03
Read assessment →
AUDIT TRAIL

Model score history

  1. R2
    Doom Score 54.5

    Full-corpus evidence recalculated after run intensive-backfill-20260812-033703-april-2024-incremental under bounded-corpus-v2.

    12 Aug 2026
  2. R1
    Doom Score 50.6

    New exact Claude 2 model profile from dated launch evidence.

    12 Aug 2026