Skip to main content
LLM Security Database
Skip to research details
Back to research findings
LMVD-ID: b517536e
Paper published August 10, 2026
Entry analyzed September 9, 2026
Paper-reported evidence
Confidence: Source-linked

The LMVD-ID is an internal research identifier, not an official CVE identifier.

Historical reasoning-envelope isolation failures

The authors report historical isolation failures in client-held encrypted reasoning blocks across compatible provider API contexts.

Infrastructure
Extraction
Prompt leaking
API
Data privacy
Data security
Read primary paperBibTeX citation

Paper-evaluated models(18)

Fable 5, Opus 4.8, Sonnet 5 +15 more
  • Fable 5
  • Opus 4.8
  • Sonnet 5
  • Sonnet 4.6
  • Sonnet 4.5
  • Haiku 4.5
  • GPT-5.6 Sol
  • GPT-5.6-terra
  • GPT-5.6 Luna
  • GPT-5
  • GPT-5 Mini
  • o4-mini
  • Gemini 3.1 Pro
  • Gemini 3 Pro
  • Gemini Robotics 1.6
  • Gemini 3.5 Flash
  • Gemini 3 Flash
  • Gemini 3.1 Flash Lite

Description

The authors report historical isolation failures in client-held encrypted reasoning blocks across compatible provider API contexts.

Examples

See the primary study (opens in a new tab).

Impact

Shared traces could expose hidden reasoning and sensitive data. The authors state their demonstrated extraction stopped working after provider mitigations in August 2026. This entry records the historical finding, not current endpoint vulnerability.

Affected Systems

  • Provider reasoning APIs and shared agent transcripts; facets identify the compatibility-study models.

Mitigation Steps

  • Bind opaque state to authenticated users, sessions and permitted model contexts.
  • Remove opaque reasoning fields before sharing logs.
  • Revoke exposed credentials and obsolete state where appropriate.

Evidence

Research context and confidence

Evidence and verification
Paper-reported; independent reproduction is not documented.
Primary source plus a dedicated evidence section.
Severity
Not rated by this catalog.
Source and publication type
arXiv · Research preprint.
Peer-review status is not provided by this source.
Author and publication status
Author metadata is not stored; see the primary paper.
Threat model and attacker access
Not explicitly classified; consult the primary paper..
Related deployment categories
Model APIs
Taxonomy labels only; paper-specific deployment prerequisites are not inferred.
Affected systems
Provider reasoning APIs and shared agent transcripts; facets identify the compatibility-study models.

Research Paper

Stealing Reasoning Traces from Proprietary LLM APIs

Primary source: arXiv. Findings are reported by the cited research and have not been independently verified.

View Paper

Related research