Five records, five observation points

A complete matrix separates period, unit, collection method, access model, direct observation, inference, and known gap across legal process, OpenAI's company report, Hugging Face host telemetry, the independent investigation, and vendor remediation.

The five evidence lanes answer different questions and carry different access and attribution boundaries.

Semantic data for Five records, five observation points
RowColumnRecordState
Access modelHugging Face host telemetryDirect access to Hugging Face systems and records within the host reviewVerified
Access modelIndependent investigationIndependent investigation with host-controlled data accessCaution
Access modelLegal processState inquiry with compulsory requests for recordsCaution
Access modelOpenAI company reportFirst-party access to OpenAI systems and recordsCaution
Access modelVendor remediationVendor-authored remediation and advisory recordsCaution
Collection methodHugging Face host telemetryAffected-host logs and forensic reconstructionVerified
Collection methodIndependent investigationAnalysis of datasets OpenAI supplied, with requests for additional materialCaution
Collection methodLegal processOfficial announcement, subpoena, and multistate letterCaution
Collection methodOpenAI company reportFirst-party internal investigation and evaluationCaution
Collection methodVendor remediationVendor publication and maintained advisory indexCaution
Known gapHugging Face host telemetryRecovered records may be incomplete; sensitive indicators are generalized or redactedCaution
Known gapIndependent investigationNo direct OpenAI infrastructure query or primary-model run access; known loss and analysis errorsCaution
Known gapLegal processNo adjudicated violation, liability, or consumer harm in the cited materialsCaution
Known gapOpenAI company reportPreliminary evaluations lack public sample sizes, intervals, and independent replicationCaution
Known gapVendor remediationThe record does not establish deployed patch coverage or remediation effectivenessCaution
InferenceHugging Face host telemetryParts of the motivational account, including an effort to obtain test solutionsCaution
InferenceIndependent investigationBehavioral analysis, including origin in broader scorer-tampering researchCaution
InferenceLegal processConsumer-protection implications remain allegations under investigationCaution
InferenceOpenAI company reportAttribution and remediation effectiveness remain first-party interpretationsCaution
InferenceVendor remediationNo independent incident attribution is inferred from the advisory laneUnavailable or historical
Direct observationHugging Face host telemetryRecovered host activity, affected assets, and operational search metadataVerified
Direct observationIndependent investigationPatterns in supplied transcripts, board data, and artifactsCaution
Direct observationLegal processExistence, scope, requests, and deadlines of the legal processVerified
Direct observationOpenAI company reportInternal activity, system configuration, impact assessment, and remediation testsCaution
Direct observationVendor remediationPublished collaboration statement and advisory entriesCaution
PeriodHugging Face host telemetryObserved July 9 at 02:28 UTC through July 13 at 14:14 UTCVerified
PeriodIndependent investigationScope June 26 through July 13, focused mostly on July 7 through July 13Caution
PeriodLegal processSubpoena served August 20; investigation announced August 24; response commanded by September 14Caution
PeriodOpenAI company reportInternal chronology through the August 26 reportsCaution
PeriodVendor remediationStatement published July 27 and updated August 5; advisories current through August 25 in the ledgerCaution
UnitHugging Face host telemetryRecovered actions, clusters, systems, and affected customer datasetsVerified
UnitIndependent investigationAgents, entries, board messages and files, standalone messages, and transcriptsCaution
UnitLegal processLegal questions and 16 requests for information or documentsCaution
UnitOpenAI company reportModels, workers, systems, credentials, repositories, tasks, controls, and evaluationsCaution
UnitVendor remediationVendor statement, advisories, CVE context, and fixed-version contextCaution
Five records, five observation points A complete matrix separates period, unit, collection method, access model, direct observation, inference, and known gap across legal process, OpenAI's company report, Hugging Face host telemetry, the independent investigation, and vendor remediation. The five evidence lanes answer different questions and carry different access and attribution boundaries. Hugging Face host telemetry Independent investigation Legal process OpenAI company report Vendor remediationAccess model Direct access to Hugging Face systems and records within the host review Independent investigation with host-controlled data access State inquiry with compulsory requests for records First-party access to OpenAI systems and records Vendor-authored remediation and advisory records Collection method Affected-host logs and forensic reconstruction Analysis of datasets OpenAI supplied, with requests for additional material Official announcement, subpoena, and multistate letter First-party internal investigation and evaluation Vendor publication and maintained advisory index Known gap Recovered records may be incomplete; sensitive indicators are generalized or redacted No direct OpenAI infrastructure query or primary-model run access; known loss and analysis errors No adjudicated violation, liability, or consumer harm in the cited materials Preliminary evaluations lack public sample sizes, intervals, and independent replication The record does not establish deployed patch coverage or remediation effectiveness Inference Parts of the motivational account, including an effort to obtain test solutions Behavioral analysis, including origin in broader scorer-tampering research Consumer-protection implications remain allegations under investigation Attribution and remediation effectiveness remain first-party interpretations No independent incident attribution is inferred from the advisory lane Direct observation Recovered host activity, affected assets, and operational search metadata Patterns in supplied transcripts, board data, and artifacts Existence, scope, requests, and deadlines of the legal process Internal activity, system configuration, impact assessment, and remediation tests Published collaboration statement and advisory entries Period Observed July 9 at 02:28 UTC through July 13 at 14:14 UTC Scope June 26 through July 13, focused mostly on July 7 through July 13 Subpoena served August 20; investigation announced August 24; response commanded by September 14 Internal chronology through the August 26 reports Statement published July 27 and updated August 5; advisories current through August 25 in the ledger Unit Recovered actions, clusters, systems, and affected customer datasets Agents, entries, board messages and files, standalone messages, and transcripts Legal questions and 16 requests for information or documents Models, workers, systems, credentials, repositories, tasks, controls, and evaluations Vendor statement, advisories, CVE context, and fixed-version context Cells use text, pattern, and outline in addition to color.
Sources
Retrieved
Units
qualitative scope fields
Transformations
  • Keep named units inside their source cells and do not aggregate across columns.
  • Preserve each source lane's terminology and attribution without assigning a score.
  • Render every missing or bounded detail as an explicit gap rather than an inferred absence.
Uncertainty
Cells summarize the reviewed source ledger. Source access and publication dates differ, and MR-2 mirrors the joint report rather than adding another investigation.
What this figure does not prove
The matrix does not rank credibility, merge incompatible units, resolve disputed interpretations, or convert allegations into findings.