The Information Machine

@AnthropicAI Would love to see fresh benchmarks after all the safeguards. 🤔

Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-06-30

Rohan Paul publicly asks Anthropic to release updated benchmarks reflecting model performance after recent safety safeguard additions, implying current published scores may no longer be accurate.

Open original ↗

Appears in

Extraction

Topics: anthropicai-benchmarksai-safetymodel-transparency

Claims

  • Recent safety safeguard updates to Anthropic models may have affected benchmark performance in ways not yet reflected in published scores.

Key quotes

@AnthropicAI Would love to see fresh benchmarks after all the safeguards. 🤔