TechCrunch · Jul 1, 2024
Kyle Wiggers argues in TechCrunch that Anthropic’s benchmark program raises questions of trust
Markdown · JSON · Original source
Kyle Wiggers’s TechCrunch article, published July 1, 2024, reports that is offering funding to develop new evaluations of AI models. The proposed would examine security and societal risks, as well as AI’s potential in scientific study, multilingual conversation and bias mitigation. Anthropic says it wants the work to help make comprehensive AI evaluation an industry standard; the article notes that its plans include evaluations aligned with the company’s own AI safety classifications. Wiggers argues that better benchmarks could address tests that fail to reflect ordinary use or may no longer measure what they claim. But he also questions whether researchers would trust a program funded by an AI vendor with commercial ambitions. Applicants may have to accept definitions of “safe” or “risky” AI they do not share, and some experts dispute the emphasis on catastrophic risks, arguing it can distract from present regulatory concerns such as . The article ends on whether independent benchmark efforts would join forces with a vendor whose loyalty, Wiggers writes, ultimately lies with shareholders. Anthropic does not explain what its proposed national-security “early warning system” would entail.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
TechCrunch · Article
Source details
Original source date: Jul 1, 2024 (2024-07-01).
Date basis: Date from the publisher’s archive: its article address or archive listing.
Added to Escape Meridian: 2026-11-18T12:08:11.277Z. First observed: 2026-10-29T07:34:10.104Z. Retrieved: 2026-10-29T07:35:35.920Z.
Referenced by
Recorded citations in public readings. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 29, 2026; originally published July 1, 2024. does not explain in the article what its proposed national-security “early warning system” would entail. The article reports that an Anthropic spokesperson declined to provide further details about the program’s funding options.
Open question
Would independent efforts join ’s program, and how would its safety classifications shape the evaluations it funds?
Source-stated
TechCrunch reports that Anthropic launched a funding program for new evaluations of AI models.
Anthropic is launching a program to fund the development of new types of benchmarks capable of evaluating the performance and impact of AI models
Source-stated
Anthropic says its program aims to help make comprehensive AI evaluation an industry standard.
“a catalyst for progress towards a future where comprehensive AI evaluation is an industry standard.”
Source-stated
Wiggers argues the program may face trust concerns because Anthropic is an AI vendor with commercial ambitions.
But given the company’s commercial ambitions in the AI race, it might be a tough one to completely trust.
Source-stated
The article says applicants may have to accept safety definitions they disagree with.
it may also force applicants to the program into accepting definitions of “safe” or “risky” AI that they might not agree with.
Source-stated
The article reports that some experts argue catastrophic-risk claims can distract from present regulatory concerns, including hallucinations.
Claims of imminent “superintelligence” serve only to draw attention away from the pressing AI regulatory issues of the day, like AI’s hallucinatory tendencies, these experts add.
Source-stated
Wiggers closes by questioning whether independent benchmark efforts would collaborate with an AI vendor whose loyalty he says lies with shareholders.
But it remains to be seen whether those efforts are willing to join forces with an AI vendor whose loyalty ultimately lies with shareholders.
Source and contributor context
TechCrunch · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Hallucination
When an AI system states something false or made up as if it were true.
Anthropic
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Editorial provenance
Version 1. Publication ID: 4c8e0e44-f9ce-786f-8d6b-88e598ee7b30.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-18T12:04:34.308Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: 3285101ace790b60ca93f046ed07d2b604b597328153901f0d012c1049ce1eab. Source text hash: 1dfa0504f0cbcdfa114608f846d5aa3ebdbeccd0f02bd7469ed99fa4dff4da66.
Revision and source-check history
Version 1 · 2026-11-18T12:08:11.277Z
Kyle Wiggers argues in TechCrunch that Anthropic’s benchmark program raises questions of trust
Kyle Wiggers’s TechCrunch article, published July 1, 2024, reports that is offering funding to develop new evaluations of AI models. The proposed would examine security and societal risks, as well as AI’s potential in scientific study, multilingual conversation and bias mitigation. Anthropic says it wants the work to help make comprehensive AI evaluation an industry standard; the article notes that its plans include evaluations aligned with the company’s own AI safety classifications. Wiggers argues that better benchmarks could address tests that fail to reflect ordinary use or may no longer measure what they claim. But he also questions whether researchers would trust a program funded by an AI vendor with commercial ambitions. Applicants may have to accept definitions of “safe” or “risky” AI they do not share, and some experts dispute the emphasis on catastrophic risks, arguing it can distract from present regulatory concerns such as . The article ends on whether independent benchmark efforts would join forces with a vendor whose loyalty, Wiggers writes, ultimately lies with shareholders. Anthropic does not explain what its proposed national-security “early warning system” would entail.
Original source
TechCrunch · Article
Source details
Original source date: Jul 1, 2024 (2024-07-01).
Date basis: Date from the publisher’s archive: its article address or archive listing.
Added to Escape Meridian: 2026-11-18T12:08:11.277Z. First observed: 2026-10-29T07:34:10.104Z. Retrieved: 2026-10-29T07:35:35.920Z.
Limits
Text as retrieved on October 29, 2026; originally published July 1, 2024. does not explain in the article what its proposed national-security “early warning system” would entail. The article reports that an Anthropic spokesperson declined to provide further details about the program’s funding options.
Open question
Would independent efforts join ’s program, and how would its safety classifications shape the evaluations it funds?
Source-stated
TechCrunch reports that Anthropic launched a funding program for new evaluations of AI models.
Anthropic is launching a program to fund the development of new types of benchmarks capable of evaluating the performance and impact of AI models
Source-stated
Anthropic says its program aims to help make comprehensive AI evaluation an industry standard.
“a catalyst for progress towards a future where comprehensive AI evaluation is an industry standard.”
Source-stated
Wiggers argues the program may face trust concerns because Anthropic is an AI vendor with commercial ambitions.
But given the company’s commercial ambitions in the AI race, it might be a tough one to completely trust.
Source-stated
The article says applicants may have to accept safety definitions they disagree with.
it may also force applicants to the program into accepting definitions of “safe” or “risky” AI that they might not agree with.
Source-stated
The article reports that some experts argue catastrophic-risk claims can distract from present regulatory concerns, including hallucinations.
Claims of imminent “superintelligence” serve only to draw attention away from the pressing AI regulatory issues of the day, like AI’s hallucinatory tendencies, these experts add.
Source-stated
Wiggers closes by questioning whether independent benchmark efforts would collaborate with an AI vendor whose loyalty he says lies with shareholders.
But it remains to be seen whether those efforts are willing to join forces with an AI vendor whose loyalty ultimately lies with shareholders.
Source and contributor context
TechCrunch · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Hallucination
When an AI system states something false or made up as if it were true.
Anthropic
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Editorial provenance
Version 1. Publication ID: 4c8e0e44-f9ce-786f-8d6b-88e598ee7b30.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-18T12:04:34.308Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: 3285101ace790b60ca93f046ed07d2b604b597328153901f0d012c1049ce1eab. Source text hash: 1dfa0504f0cbcdfa114608f846d5aa3ebdbeccd0f02bd7469ed99fa4dff4da66.
passed · 2026-11-18T12:04:34.308Z · anthropic/claude-opus-5.5
Publication: 4c8e0e44-f9ce-786f-8d6b-88e598ee7b30
attribution — pass: This is a single TechCrunch article by Kyle Wiggers, the AI Editor. The reading attributes the program details and quotations to Anthropic, the trust critique to Wiggers, and the catastrophic-risk objection to experts cited in the article. The Most Popular sidebar is page furniture, not a listing that replaces the article.
But given the company’s commercial ambitions in the AI race, it might be a tough one to completely trust.
summary — pass: The reading leads with the actor (Anthropic), the development (funding for new evaluations) and Wiggers's critique. The page defines benchmark and hallucination. It keeps the scope of the proposed tests, the aim of an industry standard, the concern about alignment with Anthropic's safety classifications, the expert dissent and the closing shareholder-loyalty line, so the author's force is preserved. The title and question follow from the evidence.
it remains to be seen whether those efforts are willing to join forces with an AI vendor whose loyalty ultimately lies with shareholders
date — pass: The evidence date is July 1, 2024 at day precision, and the reading states it at that precision. The article's "Unveiled on Monday" is consistent, because July 1, 2024 was a Monday. Later sidebar items, such as the Most Popular stories and the bio note about June 2025, are not used. The page is a publisher page, so no archive disclosure is needed. There is no hindsight framing.
Unveiled on Monday, Anthropic’s program will dole out payments to third-party organizations
support — pass: Each claim matches a verbatim excerpt. The national-security early-warning gap comes from the article's statement that the blog post does not reveal what the system entails. The spokesperson limitation is directly stated. The question about independent participation and the influence of the classifications follows from the article's closing doubts. No unsupported benefits or costs are added.
although it doesn’t reveal in the blog post what such a system might entail
media — pass: No media is selected, and no verified media candidates were supplied. The reading invents no media links.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceWhen an AI system states something false or made up as if it were true.
Definition referenceAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceWhen an AI system states something false or made up as if it were true.
Definition referenceAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore Anthropic