Skip to content

Benchmark checker Safety and security

Is MASK still meaningful?

Near ceiling Checked against the report, 5 November 2026

Fading. Top scores sit near 96, partly because evasive answers count as honest.

Why near ceiling: The top honesty score is 96.28, and evasive answers count as honest.^{[1][1] MASK leaderboard Scale AI Primary}

What “Near ceiling” means

Known problems

Evasive answers count as honest and make up 25–35% of answers.^{[1][1] MASK leaderboard Scale AI Primary}

What to look at instead

The report names no direct successor.

Where it appears in the report

Chapter 1, Benchmarks

Table 17. Honesty, sycophancy and bias benchmarks, November 2026 in “Go deeper: Honesty, sycophancy and bias (5)”

Chapter 5, Institutions and regulation

Labs grade themselves under their own frameworks, with outsiders in deeper roles

Sources

Copied from the chapter source lists. Each number links back to the source in its chapter.

[1]
MASK leaderboard, Scale AI. It gives honesty scores with confidence intervals. Primary labs.scale.com Chapter 1, source 131
[2]
MASK, arXiv 2503.03750. It separates honesty from accuracy. Primary arxiv.org Chapter 1, source 130