Claim · 2025-09-05 · publication · Stated in source · Supported

Anthropic's Frontier Red Team post of 5 September 2025 states that text-based uplift trials of the kind it reports are imperfect proxies for real-world scenarios, which involve further factors it names.

One recorded claim. The claim panel gives its qualification, the source passages with their place in each source, and how it was checked.

What it connects

LLMs and biorisk → articulates → Published evaluations do not establish real-world biological risk

Where it appears

Arguments about AI and biological risk →

Claim and evidence

LLMs and bioriskarticulates ↓Published evaluations do not establish real-world biological risk

Machine-checked, not reviewed by an editorStated in sourceSupported

Supporting evidence: primary · 1 source

Subject version: Anthropic Frontier Red Team post (5 September 2025) · Published 2025-09-05 · Effective date not established
Date: 2025-09-05 · publication
Role: Articulation
Register (kind of text): Nonfiction
Scope of the passage: The same sentence states that the trials show plausibility; the record does not separate the two halves.

Scope and limitsThe record documents the developer's own statement about its method; it is not evidence that the proposition is true, and it is not a record of the trial's result, which the post also gives and which this pass does not record. Only the post's statement about the method's limits is cited; no evaluation task content is recorded or paraphrased.

The evidence 01

primary sourceSupports the claim

LLMs and biorisk

Anthropic (Frontier Red Team) · Publication date unknown

Cite the original source

Cite the original document, not Selekoa. Check these details against it.

Source excerpt

“Text-based uplift trials like this are imperfect proxies for real-world scenarios—which involve additional factors like tacit knowledge, materials access, and actor persistence—but they demonstrate the underlying plausibility of AIs providing differentially superior assistance to a non-expert malicious actor.”

Location in source Chapter 04 ("Is an LLM's knowledge useful in an applied scenario?")

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-30

Source details and check notes

Authors: Anthropic (Frontier Red Team)
Version: Retrieved for the candidate pack on 2026-10-28 (SHA-256 4c66aa23afbeaa98dd37791df5b868f40db971a38a513caf211d41180f0f73d7). Post as served on 2026-10-28, marked "Published Sep 5, 2025" Retrieved again at admission on 2026-10-30 (SHA-256 63aadddb5bc3093907e35ab3d658898773a79ee5cdb168ef93367a8b5042b2cd). Its wording can change; earlier wording is not authenticated.
Canonical address: Open the canonical source ↗
Accessed: 2026-10-28
Language: English (en)
Attribution: Anthropic (Frontier Red Team), “LLMs and biorisk”.
Reuse: No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution.
Check notes: Candidate pack passage, AI and biosecurity (research/ai-bio). The quotation (38 words) matched the text retrieved for the candidate pack on 2026-10-28 and again at admission on 2026-10-30, in a fresh retrieval (SHA-256 63aadddb5bc3093907e35ab3d658898773a79ee5cdb168ef93367a8b5042b2cd; the bytes differ from the pack’s retrieval). Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. Quoted whole so that the sentence's second half is not dropped. A cross-family machine check (2026-10-30, item 120) re-read the record’s sources and found full support. No human editorial review.

Status and meaning

Editorial status. Not reviewed by an editor. An exploratory record that no editor has accepted. This status does not say whether a person has read any linked passage. Each evidence record above shows how it was checked.

Relation. A passage of the document meets the proposition’s criterion. The record documents what the text states, in its register (fiction and satire are labelled); it is not evidence that the proposition is true and implies no influence on or from another text.

Date. Publication date given on the post.

Record ID ab-arg-d3-anthropic-articulates