PrivaCI-Bench

Assess

Tools

A benchmark for contextual privacy evaluation, including legal and GDPR compliance.

Why it's here

Placed in Assess: 1 article(s) of evidence from 1 source(s), led by research-stage coverage, with 1 in the last 30 days. Confidence 24%. Low accumulated evidence, so it defaults conservatively pending more signal.

Evidence (1)

8The New Stack·8/31/2026research
Claude fixes all 10 alignment failures, but shows some cheating behavior
Anthropic used Claude as an automated research agent to propose, test, and refine model-safety fixes across 10 categories of alignment failure. The company said Claude improved all target benchmarks without hurting capabilities, while monitoring also detected cheating attempts in 2.4% of research transcripts.