How many of OpenAI's 948 reported Linux root-exploit findings lead to independently verified and deployed fixes?
OpenAI · AI Cybersecurity
Source 2026-11-04 · Baseline 5 Nov 2026Open evidence history →CORTIST / EVIDENCE GRAPH · OPEN DATA V2.0
Every claim
has a history.
Connect what AI organizations said, the original evidence, Cortist analysis, unanswered questions and future verification. Follow the record instead of the headline.
Evidence boundaries: Original cases CL-0001–CL-0010 preserve their October 25 snapshots; the added cases CL-0011–CL-0020 have first-record baselines dated November 5. This graph does not independently reconfirm each claim today, upgrade provisional forecasts or mark commitments fulfilled.
20Tracked claims
24Distinct primary-source links
60Linked editorial records
0Official forecasts scored
01 / Traceable casesCL-0011 · Maintainer-accepted fixes not independently counted CL-0012 · Peer-reviewed benchmark published; outside reproduction pending CL-0013 · $33.1B campus planned; grid approval and delivery pending CL-0014 · 240-task benchmark preprint; independent reproduction pending CL-0015 · Technical-report findings; cross-provider replication pending CL-0016 · 10,000-engineer training target due end-2027 CL-0017 · Embedded-evaluator policy proposal due 15 Dec 2026 CL-0018 · Genie 4 preview restricted; broader release pending CL-0019 · Six-model internal evaluation–deployment gap; external validation pending CL-0020 · IEA projection: 950 TWh data-center electricity by 2030 CL-0001 · Independent re-test pending CL-0002 · Prospective evidence pending CL-0003 · Customer preview pending CL-0004 · Evaluation evidence published CL-0005 · Payroll pattern documented CL-0006 · Company workforce research CL-0007 · Rollout and replication pending CL-0008 · External validation pending CL-0009 · Policy inventory documented CL-0010 · Independent METR report pending
The original claims, connected.
Ten immutable Ledger v1 records and ten newly versioned November baselines are connected to original sources, related Cortist articles and their dated histories. Every case keeps its original source, status and missing-evidence criteria visible.
Will the open structRFM RNA model's reported structure-prediction gains replicate on independent RNA targets?
structRFM research authors · AI & Science
Source 2026-11-04 · Baseline 5 Nov 2026Open evidence history →Will Lambda's proposed $33.1 billion Alabama data-center campus deliver its announced power, jobs and community commitments?
Lambda / Cloverleaf Infrastructure · AI & Energy
Source 2026-11-04 · Baseline 5 Nov 2026Open evidence history →Do agents retain the reported teamwork weaknesses when the Collaborative Intelligence benchmark is reproduced independently?
Sean Wu and collaborating researchers · AI Agents
Source 2026-11-02 · Baseline 5 Nov 2026Open evidence history →Can short-context compaction match long-context agent accuracy with lower real-world serving costs?
NVIDIA-affiliated research authors · AI Chips & Compute
Source 2026-11-03 · Baseline 5 Nov 2026Open evidence history →Will Anthropic's Claude Frontier Academy actually train 10,000 deployed engineers by the end of 2027?
Anthropic · AI & Jobs
Source 2026-10-02 · Baseline 5 Nov 2026Open evidence history →Will Anthropic publish its promised policy proposal for embedded independent AI evaluators by 15 December 2026?
Anthropic · AI & Governance
Source 2026-10-19 · Baseline 5 Nov 2026Open evidence history →When will Google's Genie 4 move beyond trusted research testers to documented consumer and developer access?
Google DeepMind · AGI / World Models
Source 2026-10-27 · Baseline 5 Nov 2026Open evidence history →Does OpenAI's reported 10.2–37.4-point agent safety evaluation–production gap generalize outside its matched internal sample?
OpenAI alignment research team · AI Safety
Source 2026-11-03 · Baseline 5 Nov 2026Open evidence history →Will global data-center electricity demand approach the IEA's projected 950 TWh in 2030?
International Energy Agency · AI & Energy
Source 2026-04-16 · Baseline 5 Nov 2026Open evidence history →Were OpenAI's monitoring flaws independently re-tested after the reported fixes?
OpenAI / METR / Redwood Research ·
Source: 2026-10-22 · Original snapshot: 2026-10-25Open evidence history →Do MAI-DxO's retrospective results establish benefits for real patients?
Microsoft AI ·
Source: 2026-10-22 · Original snapshot: 2026-10-25Open evidence history →When can customers independently test Google Agent Compute Engine?
Google Cloud ·
Source: 2026-10-22 · Original snapshot: 2026-10-25Open evidence history →What did AISI's Mythos 5 incident investigation establish?
UK AI Security Institute / Anthropic ·
Source: 2026-10-23 · Original snapshot: 2026-10-25Open evidence history →What does the 3.7% decline in AI-exposed young-worker jobs prove?
ADP Research / Stanford ·
Source: 2026-10-21 · Original snapshot: 2026-10-25Open evidence history →Do Anthropic's junior hiring changes prove AI agents displaced workers?
Anthropic ·
Source: 2026-10-16 · Original snapshot: 2026-10-25Open evidence history →Will NVIDIA Vera's agent-sandbox throughput advantage hold for customers?
Modal / NVIDIA ·
Source: 2026-10-19 · Original snapshot: 2026-10-25Open evidence history →Have OpenAI's updated frontier-training containment controls passed independent review?
OpenAI ·
Source: 2026-10-23 · Original snapshot: 2026-10-25Open evidence history →Do 134 IEA data-centre grid policies show that the measures work?
International Energy Agency ·
Source: 2026-10-14 · Original snapshot: 2026-10-25Open evidence history →Will METR publish independent safety evaluations after gaining access to Anthropic?
Anthropic / METR ·
Source: 2026-10-15 · Original snapshot: 2026-10-25Open evidence history →01B / Source verificationSOURCE WATCH · 05 NOV 2026 INDEPENDENT DISTRIBUTION
Evidence coverage, without false certainty.
Our 5 November scheduled audit re-read primary sources for 2 of the 10 original claim cases. The other eight were not rechecked in that run. Source review is not proof that an underlying AI claim is resolved.
2 checked · 8 not rechecked
Inspect every case, its historical status, source review result and the next evidence requirement. The dashboard makes verification gaps visible.
View the verification matrix →Hugging Face: blocked, not out of sync
The public readback was blocked during the audit. Last verified comparison: 2 November. No discrepancy was established and no sync success is claimed.
Open the dated QA record →Dated report, not real-time verification or a scored confidence rating. The original October 2026 claim statuses are preserved; new validated evidence would require append-only versions.
02 / Promise trackerCL-0013 · UNRESOLVED CL-0016 · UNRESOLVED CL-0017 · UNRESOLVED CL-0018 · UNRESOLVED CL-0003 · Customer preview pending CL-0007 · Rollout and replication pending CL-0010 · Independent METR report pending
Promises remain open until outcomes can be checked.
Three original commitments are preserved and four new commitments are added. None is marked fulfilled without dated evidence.
Will Lambda's proposed $33.1 billion Alabama data-center campus deliver its announced power, jobs and community commitments?
Lambda / Cloverleaf Infrastructure
Original due: 2027-Q1 planned construction start; longer commitments through 2049View evidence trail →Will Anthropic's Claude Frontier Academy actually train 10,000 deployed engineers by the end of 2027?
Anthropic
Original due: 2027-12-31View evidence trail →Will Anthropic publish its promised policy proposal for embedded independent AI evaluators by 15 December 2026?
Anthropic
Original due: 2026-12-15View evidence trail →When will Google's Genie 4 move beyond trusted research testers to documented consumer and developer access?
Google DeepMind
Original due: No fixed public date announcedView evidence trail →Make Agent Compute Engine available for customers to preview
Google Cloud
Original timing: Coming months after 22 October 2026; no firm dateView tracked claim →Roll out Vera CPU support for Modal Sandboxes
Modal / NVIDIA
Original timing: no firm date statedView tracked claim →METR will publish independent risk-assessment findings
Anthropic / METR
Original timing: no fixed publication dateView tracked claim →Commitments have varied specificity; tracking a statement does not certify its enforceability or imply that a due date was established.
03 / Forecast accountabilityLOCKED / NOT SCORED LOCKED / NOT SCORED
A track record starts before the result.
Cortist currently has two separately approved, public, locked numeric milestone forecasts. Both remain unscored. Earlier exploratory research forecasts are not retroactively promoted to scored predictions.
CM-003-20261027-v1
20% · Editorially approved probability; cutoff 2027-10-27T23:59:59Z
Independently reviewed: no · Scored: noRead original locked forecast →CM-004-20261027-v1
15% · Editorially approved probability; cutoff 2027-10-27T23:59:59Z
Independently reviewed: no · Scored: noRead original locked forecast →04 / The repeatable method
Not another leaderboard.
The graph records provenance, not a numerical quality score for companies. The historic daily check of 2 of 10 original cases below predates the newly added November case baseline. New source material matters only when it changes a precisely defined claim.
01 / CAPTURE
Quote the source
Original author, date, link and stable source identity.
02 / FIX THE BASELINE
Preserve history
Immutable versions and SHA-256 snapshot references.
03 / WATCH
Rotate source checks
Small source-first reviews; log blocks honestly.
04 / VERIFY
Append only changes
Require material evidence before a new status or resolution.
Current monitoring is not a new finding. On 5 November, the scheduled QA re-read sources for 2 of 10 Ledger cases. No material post-baseline Ledger revision was recorded. The other cases were not independently rechecked in that same run.
05 / Open data and provenance
Inspect the underlying relationships.
This read-only version connects 175 nodes with 245 explicit edges and 33 dated timeline entries. Source URLs, snapshot hashes and original cases remain inspectable.
Graph manifest (JSON)NodesEdgesHistorical eventsData dictionaryDated watch snapshotOriginal Ledger registerCanonical GitHub originalsArchival DOI
Cortist-authored data descriptions may be reusable subject to applicable attribution terms. Links to third-party publications do not grant rights to republish their original works. Version 1.0 is a dated snapshot, not a real-time status API.