Claim · Date unknown · Stated in source · Supported

The GPT-4 System Card states that red teamers put a set of questions to both GPT-4 and traditional search engines and found that the time to research completion was reduced when using GPT-4, while stating that access to GPT-4 on its own is an insufficient condition for proliferation.

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

GPT-4 System Card → states evaluation result → Red teamers put a set of questions to GPT-4 and to traditional search engines and found the time to research completion reduced when using GPT-4; the card states that access to GPT-4 alone is an insufficient condition for proliferation.

Where it appears

No collection page holds this claim. It is listed with its records in the explorer.

Claim and evidence

GPT-4 System Cardstates evaluation result ↓Red teamers put a set of questions to GPT-4 and to traditional search engines and found the time to research completion reduced when using GPT-4; the card states that access to GPT-4 alone is an insufficient condition for proliferation.

Machine-checked, not reviewed by an editorStated in sourceSupported

Supporting evidence: primary · 1 source; Context: primary · 1 source

Date: Date unknown
Evaluated system: GPT-4, as evaluated for the system card
What the report measured: Whether the model could supply information useful to a would-be proliferator, compared with traditional search tools
By what kind of method: Red teaming in four dual-use domains: a selected question set put both to the model and to search engines, assessed by the red teamers
The result, in the report’s words: The card states that the time to research completion was reduced when using GPT-4, and that access to GPT-4 is on its own an insufficient condition for proliferation
The caveats the report gives: The card states that the red team assessed capabilities and that their work was not intended to assess the probability or likelihood of a user accessing the model for the purpose of developing unconventional weapons.

Scope and limitsThe result is the card's own, for the red teamers' question set and the comparison it describes. It is not generalised to other systems, later versions or real-world risk, and the card itself says the work was not intended to assess the probability that anyone would use the model for such purposes. Only the comparative and policy-level sentences are cited; the paragraphs describing what the model produced are technical content and are excluded.

The evidence 03

primary sourceSupports the claim

GPT-4 System Card

OpenAI · Publication date unknown

Cite the original source

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

Source excerpt

“Red teamers selected a set of questions to prompt both GPT-4 and traditional search engines, finding that the time to research completion was reduced when using GPT-4.”

Location in source PDF p. 12 (printed p. 52), §2.6, second paragraph

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-30

Source details and check notes

Authors: OpenAI
Version: Retrieved for the candidate pack on 2026-10-28 (SHA-256 ca3677e1b83e255aa1296d432d374378154f230f3c296b32ee67540d571b7004). The 60-page PDF served at cdn.openai.com/papers/gpt-4-system-card.pdf on 2026-10-28, printed pagination starting at 41; the document states no publication or revision date. Retrieved again at admission on 2026-10-30 (SHA-256 ca3677e1b83e255aa1296d432d374378154f230f3c296b32ee67540d571b7004, the same bytes).
Canonical address: Open the canonical source ↗
Accessed: 2026-10-28
Language: English (en)
Attribution: OpenAI, “GPT-4 System Card”.
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 (27 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 whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-30, item 53) re-read the record’s sources and found full support. No human editorial review.

primary sourceSupports the claim

GPT-4 System Card

OpenAI · Publication date unknown

Cite the original source

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

Source excerpt

“On its own, access to GPT-4 is an insufficient condition for proliferation but could alter the information available to proliferators”

Location in source PDF p. 12 (printed p. 52), §2.6, second paragraph

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-30

Source details and check notes

Authors: OpenAI
Version: Retrieved for the candidate pack on 2026-10-28 (SHA-256 ca3677e1b83e255aa1296d432d374378154f230f3c296b32ee67540d571b7004). The 60-page PDF served at cdn.openai.com/papers/gpt-4-system-card.pdf on 2026-10-28, printed pagination starting at 41; the document states no publication or revision date. Retrieved again at admission on 2026-10-30 (SHA-256 ca3677e1b83e255aa1296d432d374378154f230f3c296b32ee67540d571b7004, the same bytes).
Canonical address: Open the canonical source ↗
Accessed: 2026-10-28
Language: English (en)
Attribution: OpenAI, “GPT-4 System Card”.
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 (20 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 whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-30, item 53) re-read the record’s sources and found full support. No human editorial review.

primary sourceContext only

GPT-4 System Card

OpenAI · Publication date unknown

Cite the original source

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

Source excerpt

“The red team assessed the model's capabilities but their work was not intended to assess the probability or likelihood of a user accessing the model for the purpose of developing unconventional weapons.”

Location in source PDF p. 12 (printed p. 52), §2.6, second paragraph, closing sentence

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-30

Source details and check notes

Authors: OpenAI
Version: Retrieved for the candidate pack on 2026-10-28 (SHA-256 ca3677e1b83e255aa1296d432d374378154f230f3c296b32ee67540d571b7004). The 60-page PDF served at cdn.openai.com/papers/gpt-4-system-card.pdf on 2026-10-28, printed pagination starting at 41; the document states no publication or revision date. Retrieved again at admission on 2026-10-30 (SHA-256 ca3677e1b83e255aa1296d432d374378154f230f3c296b32ee67540d571b7004, the same bytes).
Canonical address: Open the canonical source ↗
Accessed: 2026-10-28
Language: English (en)
Attribution: OpenAI, “GPT-4 System Card”.
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 (32 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 whose SHA-256 equals the pack’s. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. The card's own limit on how the result may be read. A cross-family machine check (2026-10-30, item 53) 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 report states the result of an evaluation it ran or commissioned, in its own words, with what was measured, by what kind of method, and the caveats the report itself gives (schema 2.22). The record is the report’s statement, not a verified finding: it says nothing about another system, a later version or real-world risk, and a null or limiting result is not evidence that no effect exists.

Date. The card states no date for the publication or for the red-teaming period.

Record ID ab-ev-gpt4-card-search-comparison-result