LLMs are unreliable for direct privacy analysis but good toolmakers
Use LLMs to build analysis tools, not for direct forensic or privacy analysis.
Direct AI use is therefore unsuitable for forensic and privacy analysis. AI can, however, write reliable analysis tools when their output can be validated.
LLMs may introduce errors into schema reconstructions that are hard to detect because the schemas remain internally consistent. However, they can be effective at generating code for analysis tools if the tool's output is verifiable.
Use an LLM to write a Python script that extracts data, then manually verify the script's output against known data points.
LLMs are not reliable for direct forensic or privacy analysis of complex data formats like protobuf messages due to potential errors. However, they are valuable for creating analysis tools, provided their output can be validated against known data.
The internal consistency of AI-generated schemas can mask errors, making direct analysis risky. When AI generates tools whose outputs are verifiable, the risk is mitigated, and the AI's code-generation capabilities become highly useful.
Do not rely on LLMs for direct interpretation of sensitive data without independent verification.
- Decoding What iOS Location Services Send to Apple: An LLM Is an Unreliable Analyst but a Useful Toolmaker ↗arXiv · arXiv cs.AI · Oct 15, 2026 · quoted as shared
First seen on Subvolts Oct 15, 2026. Updated Oct 19, 2026. Spotted a problem? Tell us.
- Decoding What iOS Location Services Send to Apple: An LLM Is an Unreliable Analyst but a Useful ToolmakerAnthropic · arXiv cs.AI · Oct 15, 2026
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- Case studyUsing LLMs to build tools for iOS location data analysisResearchers used an LLM to build a tool that decoded iOS location data, overcoming LLM analysis limitations. · 60 min
- HeuristicDecompose LLM vulnerability discovery into sub-tasksBreak down LLM vulnerability discovery into distinct, measurable steps for better evaluation. · 5 min
- Anti-patternRelying on gameable benchmarks for LLM vulnerability discoveryAvoid LLM benchmarks that are susceptible to data contamination or score recall against unknowable sets. · 5 min
- ToolExtractor tool to recover protobuf schemas from Apple binariesPython tool to reliably recover protobuf schemas from Apple binaries for analysis. · 30 min
- Eval methodCheck LLM state policy recall for cross-jurisdiction substitutionVerify if an LLM provides another state's policy value when asked about a specific state. · 60 min
- HeuristicAuxiliary environment signals improve LLM authorization accuracyProviding LLMs with auxiliary environment signals can substantially improve their accuracy in authorization decisions. · 30 min
Questions people ask
- Which AI tools does this work with?
- Any current model. It was seen with in the sources.
- Where did this come from?
- arXiv (arXiv cs.AI). Subvolts collected, normalized and credited it; the link above opens the original.