The BaseEssayLessWrong
'Employee-like' ≠ Employee-level
In a LessWrong linkpost, keltan argues that Dario Amodei's essay promises independent evaluators 'employee-like' access, weaker than the 'employee-level' access in Amodei's tweet, and that carve-outs for contracts and redaction leave Anthropic wide control.
Why it matters
The same thing, for different readers.
What it means for you
Promises of outside safety checks depend on small wording choices, so the exact commitment matters more than the headline.
For lawmakers
The post argues that voluntary evaluator access with company-defined limits is weaker than mandated access, and that wording like 'should' and 'mostly' is hard to enforce.
For researchers
Outside evaluators would want to know how much of their report was redacted and why, which the post says the essay does not let them state.
The question it raises
It asks what counts as a real commitment when a company writes its own contracts and decides what is commercially sensitive.
In their own words
Every quotation was found, word for word, in the text we saved from the source. That shows the words are there; it does not make them true, or prove the speaker meant them as quoted.
From the source’s own text.
“While Dario’s tweet says “employee-level”, his essay does not. Instead, the essay talks about “employee-like access”, which is not the same thing as “employee-level access”.”
“We will have the narrow ability to redact security-sensitive, legally privileged, commercially sensitive, or third-party confidential information, but we can’t redact findings just because they are unfavorable.”
“Committing to having independent evaluators with employee-like access is a great idea, and we will do the same.”
The record
- Questions
- Are AI companies putting stronger safeguards into practice? · Who can demand evidence, delay a release or stop an unsafe deployment?
- People
- keltan (Wrote it) · Dario Amodei (About them) · Sam Altman (Mentioned)
- What we read
- We read the whole source.
- Found by
- The Radar, run of October 13, 2026, via
lesswrong(tier 1: the source’s own channel or an official record) - Cite this
- LessWrong, “'Employee-like' ≠ Employee-level”, October 12, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/lesswrong-2026-10-12-employee-like-not-employee-level