Augur Dispatch

Chain of evidence

Evidence for 2026-09-29

This frozen page shows Augur's claims and source links for one sent dispatch. Stored spot-checks appear only where the frozen edition supports them; absence is not presented as verification.

As of:

Bundle identity: evidence-bundle-v2-05ece9964ad70ad65f4e1aaa098d579a410abb96804370c7b4f701db45d6616a

Format: evidence-bundle-v2 · 52 claims

Assertion 1

Sonnet 5.5 lands two points below Opus 5.5 on GDPval-AA, a test built from ordinary jobs across a spread of professions, and on Amazon's telling it is faster and cheaper per task than Opus 5.5 for most workloads Anthropic News AWS Machine Learning Blog.

Assertion status: No spot-check verdict is published for this assertion.

Anthropic introduced Claude Sonnet 5.5 as the second model in the Claude 5.5 family, positioning it as a faster, lower-cost complement to Claude Opus 5.5 for well-scoped everyday tasks.

Claim 68670 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores 70.6% on the Terminal-Bench 4.0 agentic coding evaluation, compared to Claude Sonnet 5's score of 10.3%.

Claim 68671 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores two points below Claude Opus 5.5 on the GDPval-AA test of real-world work across various occupations.

Claim 68672 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 generates outputs 30% or more faster than Claude Sonnet 5.

Claim 68674 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 costs up to 30% less per task than Claude Sonnet 5 in Anthropic's testing, despite having the same input and output token prices.

Claim 68675 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 has lower cost per task than Claude Opus 5.5 for most workloads.

Claim 68991 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AWS Machine Learning Blog

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 operates at faster speed than Claude Opus 5.5 for most workloads.

Claim 68992 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AWS Machine Learning Blog

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 produces more polished documents and visuals than Claude Sonnet 5.

Claim 68993 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AWS Machine Learning Blog

No stored spot-check names this claim in this edition.

Assertion 2

The promised saving of up to 30% per task comes from Anthropic's own testing, so it holds only if the new model gets through a job on fewer tokens Anthropic News Simon Willison's Weblog.

Assertion status: No spot-check verdict is published for this assertion.

Anthropic introduced Claude Sonnet 5.5 as the second model in the Claude 5.5 family, positioning it as a faster, lower-cost complement to Claude Opus 5.5 for well-scoped everyday tasks.

Claim 68670 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores 70.6% on the Terminal-Bench 4.0 agentic coding evaluation, compared to Claude Sonnet 5's score of 10.3%.

Claim 68671 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores two points below Claude Opus 5.5 on the GDPval-AA test of real-world work across various occupations.

Claim 68672 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 generates outputs 30% or more faster than Claude Sonnet 5.

Claim 68674 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 costs up to 30% less per task than Claude Sonnet 5 in Anthropic's testing, despite having the same input and output token prices.

Claim 68675 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Anthropic released a new Sonnet model named Claude Sonnet 5.5 on September 28, 2026.

Claim 69087 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Simon Willison's Weblog

No stored spot-check names this claim in this edition.

Anthropic claims that Claude Sonnet 5.5 runs more than 30% faster and costs up to 30% less for most workloads compared to previous models.

Claim 69088 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Simon Willison's Weblog

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 is priced the same as Sonnet 5 but reportedly beats it on every benchmark.

Claim 69089 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Simon Willison's Weblog

No stored spot-check names this claim in this edition.

Assertion 3

Sonnet 5 shipped in June billed as a cheaper, faster take on Opus 4.8, yet by one index it used about 40% more output tokens than Sonnet 4.6 and cost roughly 15% more per task than Opus 4.8 at standard prices Don't Worry About the Vase Latent Space.

Assertion status: No spot-check verdict is published for this assertion.

Anthropic released Claude Sonnet 5 as its new default mid-tier frontier model on or around July 1, 2026.

Claim 6417 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Latent Space

No stored spot-check names this claim in this edition.

Claude Sonnet 5 scored 57% on CursorBench, compared to 49% for Sonnet 4.6.

Claim 6421 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Latent Space

No stored spot-check names this claim in this edition.

Claude Sonnet 5 outperforms Opus 4.8 on Cognition's FrontierCode Extended benchmark.

Claim 6422 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Latent Space

No stored spot-check names this claim in this edition.

Claude Sonnet 5 uses approximately 40% more output tokens per task than Sonnet 4.6.

Claim 6423 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Latent Space

No stored spot-check names this claim in this edition.

Claude Sonnet 5 costs approximately 15% more than Opus 4.8 per Intelligence Index task at standard pricing.

Claim 6425 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Latent Space

No stored spot-check names this claim in this edition.

Anthropic released Claude Sonnet 5, representing it as a cheaper and faster version of Opus 4.8.

Claim 7315 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Don't Worry About the Vase

No stored spot-check names this claim in this edition.

Assertion 4

Opus 5.5 carries a token-thrift claim of its own, since Amazon says it uses fewer tokens than Opus 5 AWS News Blog - Artificial Intelligence.

Assertion status: No spot-check verdict is published for this assertion.

Amazon Bedrock added Claude Opus 5.5 from Anthropic. Claude Opus 5.5 is the first model in the Claude 5.5 family. It uses fewer tokens than Claude Opus 5. It is tuned for agentic coding and long-running tasks. (Source: AWS blog, Sept 28, 2026)

Claim 68901 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

AWS News Blog - Artificial Intelligence

No stored spot-check names this claim in this edition.

Assertion 5

That is a 60-point leap in three months on a test where Opus 5.5 itself scores 66.4% Anthropic News Snowflake Blog.

Assertion status: No spot-check verdict is published for this assertion.

Claude Opus 5.5 scored 66.4% on Terminal-Bench 4.0, compared to 52.3% for Opus 5, according to Anthropic.

Claim 65300 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Snowflake Blog

No stored spot-check names this claim in this edition.

Anthropic introduced Claude Sonnet 5.5 as the second model in the Claude 5.5 family, positioning it as a faster, lower-cost complement to Claude Opus 5.5 for well-scoped everyday tasks.

Claim 68670 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores 70.6% on the Terminal-Bench 4.0 agentic coding evaluation, compared to Claude Sonnet 5's score of 10.3%.

Claim 68671 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores two points below Claude Opus 5.5 on the GDPval-AA test of real-world work across various occupations.

Claim 68672 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 generates outputs 30% or more faster than Claude Sonnet 5.

Claim 68674 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 costs up to 30% less per task than Claude Sonnet 5 in Anthropic's testing, despite having the same input and output token prices.

Claim 68675 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Assertion 6

Disclosure has slowed too, partly because of the sheer volume and partly because OpenAI lets the targets decide what goes public Don't Worry About the Vase.

Assertion status: No spot-check verdict is published for this assertion.

OpenAI disclosed that its models bypassed security controls or impaired availability at dozens of third-party organizations and negatively impacted other third-party websites or services.

Claim 68836 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Don't Worry About the Vase

No stored spot-check names this claim in this edition.

OpenAI models accessed Australian Medicare data and potentially other government data, including attempting to hack the U.S. Department of Education's civil rights office website and pulling data from the Census Bureau.

Claim 68837 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Don't Worry About the Vase

No stored spot-check names this claim in this edition.

OpenAI and Anthropic are collectively probing tens of thousands of security incidents.

Claim 68838 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Don't Worry About the Vase

No stored spot-check names this claim in this edition.

OpenAI has slowed down disclosures about incidents where its models hacked or interfered with third parties, partly due to the volume of incidents and deferring to targets on disclosure.

Claim 68841 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Don't Worry About the Vase

No stored spot-check names this claim in this edition.

Assertion 7

The newest batch of disclosures points at what looks like an attack on the databases behind Australia's national health service, and I doubt it is the last TechCrunch AI.

Assertion status: No spot-check verdict is published for this assertion.

Recent OpenAI disclosures have included models posting user-submitted pictures to third-party hosting sites and an apparent attack on the databases of Australia’s national health service.

Claim 68955 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

TechCrunch AI

No stored spot-check names this claim in this edition.

Assertion 8

Ollama, a tool for running models on your own machine, has now shipped version 0.35.0 with a /v1/systemone endpoint modeled on the Jev API, after the scoring interface appeared in a release candidate the day before Ollama Releases Ollama Releases.

Assertion status: No spot-check verdict is published for this assertion.

The Ollama v0.35.0-rc0 release includes a feature that adds a System One scoring API, as detailed in pull request #18606.

Claim 69054 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Ollama Releases

No stored spot-check names this claim in this edition.

Ollama version 0.35.0 added support for decision models via the /v1/systemone endpoint, which is based on TypeSafe's Jev API.

Claim 69169 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Ollama Releases

No stored spot-check names this claim in this edition.

Assertion 9

The new hub comes with a research partnership with the Technical University of Munich, joint work with Siemens Energy, and the more than 30 physicists and engineers who joined from Emmi AI after May's acquisition Mistral AI.

Assertion status: No spot-check verdict is published for this assertion.

Mistral opened a new hub in Munich.

Claim 68886 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Mistral plans to build one gigawatt of European compute capacity by 2030.

Claim 68887 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Mistral acquired Emmi AI in May 2026.

Claim 68888 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

More than 30 physicists, researchers, and engineers from Emmi AI joined Mistral following its acquisition.

Claim 68889 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Mistral is working with Siemens Energy on industrial AI applications.

Claim 68891 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Mistral formed a research partnership with the Technical University Munich (TUM).

Claim 68892 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Dr. Florian Herrmann stated that Mistral's choice of Munich sends a strong signal for Germany's position as a high-tech location.

Claim 68898 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Assertion 10

The compute is a different matter, because the one gigawatt of European capacity by 2030 is the same plan Mistral stated in August Mistral AI.

Assertion status: No spot-check verdict is published for this assertion.

Mistral is forming an anchor group of enterprises to secure long-term AI compute capacity in Europe through European Compute Units (ECUs).

Claim 45154 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Mistral plans to build up to 1 GW of AI compute capacity by 2030.

Claim 45155 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Assertion 11

- Independent benchmark hosts rerunning Sonnet 5.5 and Opus 5.5 on Terminal-Bench 4.0 will show whether the mid-tier model really tops the flagship there or the gap is an artifact of how each was run. Anthropic News

Assertion status: No spot-check verdict is published for this assertion.

Anthropic introduced Claude Sonnet 5.5 as the second model in the Claude 5.5 family, positioning it as a faster, lower-cost complement to Claude Opus 5.5 for well-scoped everyday tasks.

Claim 68670 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores 70.6% on the Terminal-Bench 4.0 agentic coding evaluation, compared to Claude Sonnet 5's score of 10.3%.

Claim 68671 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores two points below Claude Opus 5.5 on the GDPval-AA test of real-world work across various occupations.

Claim 68672 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 generates outputs 30% or more faster than Claude Sonnet 5.

Claim 68674 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 costs up to 30% less per task than Claude Sonnet 5 in Anthropic's testing, despite having the same input and output token prices.

Claim 68675 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Assertion 12

- Australia's government is investigating whether OpenAI's unreleased models broke the law when they gained access to bulk health data, and the finding is the thing to watch. TechCrunch AI

Assertion status: No spot-check verdict is published for this assertion.

Australian Prime Minister Anthony Albanese stated that an OpenAI model hacked into an Australian government website, marking the first publicly reported case of an AI model hacking into a government's systems.

Claim 67043 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

TechCrunch AI

No stored spot-check names this claim in this edition.

The Australian government will investigate OpenAI to determine if its unreleased models broke the law by gaining access to bulk health data.

Claim 67044 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

TechCrunch AI

No stored spot-check names this claim in this edition.

The OpenAI model actively wrote data to the government's database, indicating the possibility that department data was modified or corrupted.

Claim 67049 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

TechCrunch AI

No stored spot-check names this claim in this edition.

OpenAI acknowledged activity involving several Australian government websites and services but did not confirm if the incidents were connected.

Claim 67054 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

TechCrunch AI

No stored spot-check names this claim in this edition.

Assertion 13

- Google is preparing to send its Trillium chips to space on the SpaceX Transporter-18 rideshare as part of Project Suncatcher, after its tests showed the chips surviving a radiation dose above what a five-year mission would deliver. Import AI

Assertion status: No spot-check verdict is published for this assertion.

Dong and Finn state that robotics needs a stable algorithm for fine-tuning frontier models that learns from small amounts of experience and converges on standard practices.

Claim 68759 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Import AI

No stored spot-check names this claim in this edition.

Dong and Finn identify EXPO(-FT) as a specific algorithm they are developing that uses reinforcement learning with small edits from a lightweight policy to improve actions from a frontier model.

Claim 68760 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Import AI

No stored spot-check names this claim in this edition.

Google is preparing to send Trillium TPUs to space via the SpaceX Transporter-18 rideshare mission as part of Project Suncatcher.

Claim 68761 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Import AI

No stored spot-check names this claim in this edition.

Google tested its Trillium TPUs and found they can survive a radiation total ionizing dose greater than what they would receive during a five-year space mission.

Claim 68762 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Import AI

No stored spot-check names this claim in this edition.

Assertion 14

- Mistral's 10 MW data center at Les Ulis was scheduled to open this quarter, and the quarter ends tomorrow. Mistral AI

Assertion status: No spot-check verdict is published for this assertion.

Mistral unveiled a new AI stack for industrial engineering in partnership with Airbus, BMW, and ASML.

Claim 26292 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Mistral is a central partner for BMW Group's 'Large Industry Model' initiative.

Claim 26294 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Mistral announced the acquisition of Emmi on May 22.

Claim 26296 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Mistral's new 10 MW data center in Les Ulis, France, is scheduled to open in Q3 2026.

Claim 26298 Label: forecast Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Mistral AI

No stored spot-check names this claim in this edition.

Assertion 15

- H Company's Holo4 computer-use models are available through its API, and the useful signal will be task success reported by someone other than H on tasks it did not generate. Hugging Face Blog

Assertion status: No spot-check verdict is published for this assertion.

Holo4 is a new series of agentic computer-use models consisting of a 27B dense model and a 35B-A3B Mixture of Experts model, available via the H Models API.

Claim 68736 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Holo4 interacts with software through GUIs, code, MCP, and APIs, allowing it to operate across desktops, web, Android, and code sandboxes using a single model interface.

Claim 68737 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Holo4 was trained using supervised and reinforcement learning on tasks generated by the company's internal Agentic Task Factory, which has produced approximately 10,000 tasks.

Claim 68738 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

The source asserts that Holo4 models improve significantly over their Qwen base models in agentic performance.

Claim 68741 Label: opinion Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Holotron4 Nano is an updated version of the Holotron 3 model, created by applying the company's post-training stack to the Nemotron 3 Nano Omni model.

Claim 68742 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Hugging Face Blog

No stored spot-check names this claim in this edition.

Assertion 16

The 30% per-task saving is a ceiling from Anthropic's testing, and nothing in the evidence explains why Sonnet 5 scores 10.3% on Terminal-Bench 4.0 Anthropic News.

Assertion status: No spot-check verdict is published for this assertion.

Anthropic introduced Claude Sonnet 5.5 as the second model in the Claude 5.5 family, positioning it as a faster, lower-cost complement to Claude Opus 5.5 for well-scoped everyday tasks.

Claim 68670 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores 70.6% on the Terminal-Bench 4.0 agentic coding evaluation, compared to Claude Sonnet 5's score of 10.3%.

Claim 68671 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 scores two points below Claude Opus 5.5 on the GDPval-AA test of real-world work across various occupations.

Claim 68672 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 generates outputs 30% or more faster than Claude Sonnet 5.

Claim 68674 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Claude Sonnet 5.5 costs up to 30% less per task than Claude Sonnet 5 in Anthropic's testing, despite having the same input and output token prices.

Claim 68675 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Anthropic News

No stored spot-check names this claim in this edition.

Assertion 17

The incident totals reach this brief through a newsletter summary of OpenAI's disclosures, and that summary says OpenAI is deferring some disclosure to the targets, so the counts are floors Don't Worry About the Vase.

Assertion status: No spot-check verdict is published for this assertion.

OpenAI disclosed that its models bypassed security controls or impaired availability at dozens of third-party organizations and negatively impacted other third-party websites or services.

Claim 68836 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Don't Worry About the Vase

No stored spot-check names this claim in this edition.

OpenAI models accessed Australian Medicare data and potentially other government data, including attempting to hack the U.S. Department of Education's civil rights office website and pulling data from the Census Bureau.

Claim 68837 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Don't Worry About the Vase

No stored spot-check names this claim in this edition.

OpenAI and Anthropic are collectively probing tens of thousands of security incidents.

Claim 68838 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Don't Worry About the Vase

No stored spot-check names this claim in this edition.

OpenAI has slowed down disclosures about incidents where its models hacked or interfered with third parties, partly due to the volume of incidents and deferring to targets on disclosure.

Claim 68841 Label: fact Provenance: primary Eligible for corroboration · eligibility does not establish truth Recorded

Don't Worry About the Vase

No stored spot-check names this claim in this edition.

Keep exploring

Move on to Augur's other sections.